[
  {
    "id": "2609.14261v1",
    "title": "VGFM: Expressive Robot Policies via Dense Value Guidance in Flow Matching",
    "abstract": "Recent robot learning paradigms increasingly rely on large offline datasets of robotic interactions to train control policies. Expressive generative models enable rich and multimodal action representations, expanding the capability of this paradigm for complex robotic control. However, policy improvement with multi-step generative actors remains challenging. In offline reinforcement learning (RL), incorporating value-based objectives along generative trajectories often introduces substantial training complexity, including backpropagation through time (BPTT), auxiliary architectures, or distillation losses. We propose Value-Guided Flow Matching (VGFM), a scalable offline RL framework that enables dense value-guided shaping within a flow-based policy while avoiding BPTT and additional algorithmic overhead. VGFM parameterizes the policy as a conditional flow-matching model in action (x-prediction) space, ensuring that each intermediate flow step produces a valid robot action that can be directly evaluated by a standard offline RL critic. This design allows value guidance to be applied at randomly sampled flow times without differentiating through the entire generative trajectory, while preserving inference-time flexibility by varying the discretization of the underlying flow ODE without retraining. Evaluated on robotic locomotion and manipulation tasks in OGBench, VGFM achieves strong performance across a wide range of tasks under rigorous evaluation protocols. With minimal hyperparameter tuning, these results demonstrate that VGFM provides a simple, scalable, and effective approach for expressive policy learning in long-horizon, goal-oriented robotic control.",
    "history": "Submission history From: Prajwal Koirala [ view email ] [v1] Sun, 13 Sep 2026 03:50:52 UTC (2,734 KB)",
    "meta": {
      "citation_title": "VGFM: Expressive Robot Policies via Dense Value Guidance in Flow Matching",
      "citation_author": "Campbell, Mark",
      "citation_date": "2026/09/13",
      "citation_online_date": "2026/09/13",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.14261",
      "citation_arxiv_id": "2609.14261",
      "citation_abstract": "Recent robot learning paradigms increasingly rely on large offline datasets of robotic interactions to train control policies. Expressive generative models enable rich and multimodal action representations, expanding the capability of this paradigm for complex robotic control. However, policy improvement with multi-step generative actors remains challenging. In offline reinforcement learning (RL), incorporating value-based objectives along generative trajectories often introduces substantial training complexity, including backpropagation through time (BPTT), auxiliary architectures, or distillation losses. We propose Value-Guided Flow Matching (VGFM), a scalable offline RL framework that enables dense value-guided shaping within a flow-based policy while avoiding BPTT and additional algorithmic overhead. VGFM parameterizes the policy as a conditional flow-matching model in action (x-prediction) space, ensuring that each intermediate flow step produces a valid robot action that can be directly evaluated by a standard offline RL critic. This design allows value guidance to be applied at randomly sampled flow times without differentiating through the entire generative trajectory, while preserving inference-time flexibility by varying the discretization of the underlying flow ODE without retraining. Evaluated on robotic locomotion and manipulation tasks in OGBench, VGFM achieves strong performance across a wide range of tasks under rigorous evaluation protocols. With minimal hyperparameter tuning, these results demonstrate that VGFM provides a simple, scalable, and effective approach for expressive policy learning in long-horizon, goal-oriented robotic control."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-13T11:50:52+08:00"
  },
  {
    "id": "2609.13851v1",
    "title": "ReWeight: Leveraging Human Data for VLA Post-Training via Demonstration Retrieval and Sample Weighting",
    "abstract": "Post-training vision-language-action (VLA) models for specific robots and tasks requires in-domain demonstrations, yet collecting diverse robot data is costly. Egocentric human demonstrations provide a scalable alternative, but directly mixing human and robot data can introduce cross-embodiment discrepancies and degrade policy performance. To address this challenge, we introduce ReWeight, a framework that incorporates human data into VLA post-training through demonstration-level retrieval and sample-level weighting. ReWeight learns a cross-embodiment visuomotor representation that combines visual observations with future actions to measure behavioral similarity between human and robot demonstrations. Based on optimal transport, it retrieves human demonstrations relevant to the target robot data and assigns larger weights to samples with smaller cross-embodiment discrepancies. We evaluate ReWeight using $\\pi_{0.5}$ across eight simulation tasks and four real-world tasks under both clean and randomized settings. In simulation, ReWeight improves the average success rate of post-trained $\\pi_{0.5}$ from 39% with only robot data and 44% with randomly mixed human-robot data to 57%. In the physical experimental setting, it achieves an average success rate of 68.8%, outperforming the baselines by 28.8% and 13.8%, respectively. Overall, ReWeight provides an effective paradigm for transforming abundant egocentric human experience into transferable supervision for robot learning. (Project webpage: this https URL )",
    "history": "Submission history From: Chenwei Wang [ view email ] [v1] Sat, 12 Sep 2026 10:09:06 UTC (6,521 KB)",
    "meta": {
      "citation_title": "ReWeight: Leveraging Human Data for VLA Post-Training via Demonstration Retrieval and Sample Weighting",
      "citation_author": "Jiang, Zhongliang",
      "citation_date": "2026/09/12",
      "citation_online_date": "2026/09/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13851",
      "citation_arxiv_id": "2609.13851",
      "citation_abstract": "Post-training vision-language-action (VLA) models for specific robots and tasks requires in-domain demonstrations, yet collecting diverse robot data is costly. Egocentric human demonstrations provide a scalable alternative, but directly mixing human and robot data can introduce cross-embodiment discrepancies and degrade policy performance. To address this challenge, we introduce ReWeight, a framework that incorporates human data into VLA post-training through demonstration-level retrieval and sample-level weighting. ReWeight learns a cross-embodiment visuomotor representation that combines visual observations with future actions to measure behavioral similarity between human and robot demonstrations. Based on optimal transport, it retrieves human demonstrations relevant to the target robot data and assigns larger weights to samples with smaller cross-embodiment discrepancies. We evaluate ReWeight using $\\pi_{0.5}$ across eight simulation tasks and four real-world tasks under both clean and randomized settings. In simulation, ReWeight improves the average success rate of post-trained $\\pi_{0.5}$ from 39% with only robot data and 44% with randomly mixed human-robot data to 57%. In the physical experimental setting, it achieves an average success rate of 68.8%, outperforming the baselines by 28.8% and 13.8%, respectively. Overall, ReWeight provides an effective paradigm for transforming abundant egocentric human experience into transferable supervision for robot learning. (Project webpage: https://reweight-vla.github.io/)"
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T18:09:06+08:00"
  },
  {
    "id": "2609.13845v1",
    "title": "LePlanner: An Iterative Amortized Controller For World Models",
    "abstract": "World models trained with joint-embedding predictive architectures learn compact, structured latent representations from physical interaction, yet planning in these latent spaces typically relies on one of two costly approaches. Search-based planners such as CEM, MPPI, and iCEM optimize action sequences through many predictor rollouts, achieving strong performance at the cost of high per-decision compute and latency. Policy-based methods amortize inference into a single forward pass but can degrade on contact-rich tasks where the demonstration distribution is multimodal. We propose LePlanner, an amortized iterative controller that learns to construct and refine latent action sequences through a frozen world-model predictor. LePlanner is trained with an arrival-and-hold objective that encourages the controller to reach the goal at the earliest feasible horizon and remain there. This addresses horizon-reset procrastination, a failure mode in which repeated receding-horizon replanning continually postpones goal arrival. An additional action-Gaussian loss keeps generated actions near the support of the offline dataset. Across navigation, contact-rich manipulation, and continuous-control environments, LePlanner matches or exceeds search-based planners while requiring an order of magnitude fewer predictor evaluations and 3-49x lower wall-clock time per decision. It achieves success rates of 98% on PushT, 100% on Reacher, 100% on TwoRooms, and 92% on the OGBench Cube task. These results show that much of the structure discovered through online search can be amortized into a lightweight iterative policy, enabling fast, horizon-aware, nonlinear physical control without online optimization.",
    "history": "Submission history From: Saksham Bansal [ view email ] [v1] Sat, 12 Sep 2026 10:01:21 UTC (3,080 KB)",
    "meta": {
      "citation_title": "LePlanner: An Iterative Amortized Controller For World Models",
      "citation_author": "M, Vrishin",
      "citation_date": "2026/09/12",
      "citation_online_date": "2026/09/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13845",
      "citation_arxiv_id": "2609.13845",
      "citation_abstract": "World models trained with joint-embedding predictive architectures learn compact, structured latent representations from physical interaction, yet planning in these latent spaces typically relies on one of two costly approaches. Search-based planners such as CEM, MPPI, and iCEM optimize action sequences through many predictor rollouts, achieving strong performance at the cost of high per-decision compute and latency. Policy-based methods amortize inference into a single forward pass but can degrade on contact-rich tasks where the demonstration distribution is multimodal. We propose LePlanner, an amortized iterative controller that learns to construct and refine latent action sequences through a frozen world-model predictor. LePlanner is trained with an arrival-and-hold objective that encourages the controller to reach the goal at the earliest feasible horizon and remain there. This addresses horizon-reset procrastination, a failure mode in which repeated receding-horizon replanning continually postpones goal arrival. An additional action-Gaussian loss keeps generated actions near the support of the offline dataset. Across navigation, contact-rich manipulation, and continuous-control environments, LePlanner matches or exceeds search-based planners while requiring an order of magnitude fewer predictor evaluations and 3-49x lower wall-clock time per decision. It achieves success rates of 98% on PushT, 100% on Reacher, 100% on TwoRooms, and 92% on the OGBench Cube task. These results show that much of the structure discovered through online search can be amortized into a lightweight iterative policy, enabling fast, horizon-aware, nonlinear physical control without online optimization."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T18:01:21+08:00"
  },
  {
    "id": "2609.13695v1",
    "title": "GROOVE: Geometry-Guided Reduction of Operational-Space Jerk in VLA Execution",
    "abstract": "Chunked vision language action (VLA) policies execute several commands per query, but jerk within chunks and across replanning boundaries can induce oscillatory motion and sharp actuator transients. We present GROOVE, an online regulator that searches directional correction regions around the raw three dimensional end effector (EEF) path, without retraining or additional VLA inference. It optimizes the new chunk using delivered commands as boundary conditions, reducing boundary and within chunk jerk while bounding cumulative translation and local axis angle deviation from the raw plan after every command. Using quadratic programs (QPs), GROOVE generates a cube reference and thirteen directional candidates, then selects the one with the lowest command space jerk under a reference relative deviation cap. On a held out LIBERO benchmark, GROOVE achieves the largest reductions among the evaluated methods, reducing translational and rotational EEF jerk by 33.02% and 43.42%, respectively, with task success of 95.75% versus 93.75% for raw execution. Across 50 matched UR5e pairs with measured execution timing, it reduces translational and rotational tool center point (TCP) jerk by 16.39% and 19.49% and joint current slew by 29.09%.",
    "history": "Submission history From: Sangho Yun [ view email ] [v1] Sat, 12 Sep 2026 04:05:17 UTC (1,613 KB)",
    "meta": {
      "citation_title": "GROOVE: Geometry-Guided Reduction of Operational-Space Jerk in VLA Execution",
      "citation_author": "Yu, Hwanjo",
      "citation_date": "2026/09/12",
      "citation_online_date": "2026/09/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13695",
      "citation_arxiv_id": "2609.13695",
      "citation_abstract": "Chunked vision language action (VLA) policies execute several commands per query, but jerk within chunks and across replanning boundaries can induce oscillatory motion and sharp actuator transients. We present GROOVE, an online regulator that searches directional correction regions around the raw three dimensional end effector (EEF) path, without retraining or additional VLA inference. It optimizes the new chunk using delivered commands as boundary conditions, reducing boundary and within chunk jerk while bounding cumulative translation and local axis angle deviation from the raw plan after every command. Using quadratic programs (QPs), GROOVE generates a cube reference and thirteen directional candidates, then selects the one with the lowest command space jerk under a reference relative deviation cap. On a held out LIBERO benchmark, GROOVE achieves the largest reductions among the evaluated methods, reducing translational and rotational EEF jerk by 33.02% and 43.42%, respectively, with task success of 95.75% versus 93.75% for raw execution. Across 50 matched UR5e pairs with measured execution timing, it reduces translational and rotational tool center point (TCP) jerk by 16.39% and 19.49% and joint current slew by 29.09%."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T12:05:17+08:00"
  },
  {
    "id": "2609.13651v1",
    "title": "Compositional Shift Algebra: Extrapolating Mixed Robot Shifts Without Mixed Finetuning",
    "abstract": "Robot deployments rarely change one mechanism at a time: cameras, action interfaces, and physical dynamics often shift together. Prior adaptation recipes either finetune a new model for every mix or attempt to select which module to update. We instead learn shift operators on a modular stack z{=}E(o), a{=}g(z,u), z'{=}f(z,a) and compose them. Compositional Shift Algebra (CSA) fits single-factor observation, policy, and dynamics operators from exact-reset probes, then extrapolates held-out mixed shifts by operator composition---without mixed-shift finetuning. On ManiSkill StackCube, residual CSA matches an oracle mixed inverse on held-out mixes (success 1.0 over 10 seeds) while beating best-single / zero-shot / parameter-average baselines by { approx}67 pp. RGB-D vision-in-the-loop composition remains near oracle and far above non-compositional arms; a delay commutator stress shows ordered necessity for policy timesdelay. On a second task (PickCube), residual CSA again reaches compose 1.0 vs. 0.33 non-compositional (n{=}10), and an L1 vision controller without privileged cube/goal poses or grasp flags in the control loop retains compose 0.95 vs. 0.00. Main-track upgrades freeze PushCube (+33 pp), PegInsertion joint8 / pose7 EE (+67 pp each), and thin BC under frozen CSA (+67 pp); deeper BC and fair adapt baselines still need compose (+67 pp each), vision-localized BC needs compose (+56 pp), and delay favors ordered/few-shot deploy. We report Intervention-Gated Adaptation as a negative control.",
    "history": "Submission history From: Jinting Hang [ view email ] [v1] Sat, 12 Sep 2026 02:14:26 UTC (102 KB)",
    "meta": {
      "citation_title": "Compositional Shift Algebra: Extrapolating Mixed Robot Shifts Without Mixed Finetuning",
      "citation_author": "Cai, Zhenhui",
      "citation_date": "2026/09/12",
      "citation_online_date": "2026/09/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13651",
      "citation_arxiv_id": "2609.13651",
      "citation_abstract": "Robot deployments rarely change one mechanism at a time: cameras, action interfaces, and physical dynamics often shift together. Prior adaptation recipes either finetune a new model for every mix or attempt to select which module to update. We instead learn shift operators on a modular stack z{=}E(o), a{=}g(z,u), z'{=}f(z,a) and compose them. Compositional Shift Algebra (CSA) fits single-factor observation, policy, and dynamics operators from exact-reset probes, then extrapolates held-out mixed shifts by operator composition---without mixed-shift finetuning. On ManiSkill StackCube, residual CSA matches an oracle mixed inverse on held-out mixes (success 1.0 over 10 seeds) while beating best-single / zero-shot / parameter-average baselines by { approx}67 pp. RGB-D vision-in-the-loop composition remains near oracle and far above non-compositional arms; a delay commutator stress shows ordered necessity for policy timesdelay. On a second task (PickCube), residual CSA again reaches compose 1.0 vs. 0.33 non-compositional (n{=}10), and an L1 vision controller without privileged cube/goal poses or grasp flags in the control loop retains compose 0.95 vs. 0.00. Main-track upgrades freeze PushCube (+33 pp), PegInsertion joint8 / pose7 EE (+67 pp each), and thin BC under frozen CSA (+67 pp); deeper BC and fair adapt baselines still need compose (+67 pp each), vision-localized BC needs compose (+56 pp), and delay favors ordered/few-shot deploy. We report Intervention-Gated Adaptation as a negative control."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T10:14:26+08:00"
  },
  {
    "id": "2609.13053v1",
    "title": "Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model",
    "abstract": "Visual goal and dynamics prediction can provide language-conditioned robot policies with both a target outcome and a representation of action-dependent scene changes. We bring these predictions into action generation and selection through a shared trajectory model. Dynin-Robotics implements this formulation on Dynin-Omni, an omnimodal masked-diffusion backbone, representing language, visual observations, goals, and actions as discrete tokens. By varying conditioning and target spans, the same model learns action prediction, action-conditioned next-observation prediction, terminal goal-state prediction, and trajectory-to-instruction reconstruction. These interfaces support test-time scaling through goal prediction, action-candidate evaluation, and joint refinement of action and future-state predictions. We continually pretrain the model on approximately 1.33 million trajectories from 48 Open X-Embodiment datasets and adapt it separately to downstream domains. On two VLABench tasks, robot pretraining improves adaptation within a fixed Stage-2 step budget, and the full objective mixture improves shifted-instruction success over Policy-only post-training under the same coupled decoder. Combining goal guidance with joint action-next-state denoising further improves shifted-instruction success over action-only decoding; the benefit depends on how the predictions are composed. Dynin-Robotics achieves competitive performance on LIBERO and zero-shot LIBERO-Plus, together with a 78.4% average success rate across four manipulation conditions on a Franka Research 3 robot. An optimized block-parallel implementation accelerates model-side action decoding by up to 29.2x relative to the base implementation under the reported profiling setup. These results support shared trajectory modeling as a common interface for learning complementary robot objectives and composing their predictions during control.",
    "history": "Submission history From: Hoeun Lee [ view email ] [v1] Fri, 11 Sep 2026 16:49:26 UTC (40,074 KB)",
    "meta": {
      "citation_title": "Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model",
      "citation_author": "Do, Jaeyoung",
      "citation_date": "2026/09/11",
      "citation_online_date": "2026/09/11",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13053",
      "citation_arxiv_id": "2609.13053",
      "citation_abstract": "Visual goal and dynamics prediction can provide language-conditioned robot policies with both a target outcome and a representation of action-dependent scene changes. We bring these predictions into action generation and selection through a shared trajectory model. Dynin-Robotics implements this formulation on Dynin-Omni, an omnimodal masked-diffusion backbone, representing language, visual observations, goals, and actions as discrete tokens. By varying conditioning and target spans, the same model learns action prediction, action-conditioned next-observation prediction, terminal goal-state prediction, and trajectory-to-instruction reconstruction. These interfaces support test-time scaling through goal prediction, action-candidate evaluation, and joint refinement of action and future-state predictions. We continually pretrain the model on approximately 1.33 million trajectories from 48 Open X-Embodiment datasets and adapt it separately to downstream domains. On two VLABench tasks, robot pretraining improves adaptation within a fixed Stage-2 step budget, and the full objective mixture improves shifted-instruction success over Policy-only post-training under the same coupled decoder. Combining goal guidance with joint action-next-state denoising further improves shifted-instruction success over action-only decoding; the benefit depends on how the predictions are composed. Dynin-Robotics achieves competitive performance on LIBERO and zero-shot LIBERO-Plus, together with a 78.4% average success rate across four manipulation conditions on a Franka Research 3 robot. An optimized block-parallel implementation accelerates model-side action decoding by up to 29.2x relative to the base implementation under the reported profiling setup. These results support shared trajectory modeling as a common interface for learning complementary robot objectives and composing their predictions during control."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T00:49:26+08:00"
  },
  {
    "id": "2609.12898v1",
    "title": "UniPart: Towards Zero-shot Language-Grounded 3D Part Segmentation for Embodied Interaction",
    "abstract": "Fine-grained robotic manipulation depends on understanding parts, not only whole objects. Existing 3D foundation models tend to be either generalized but object-aware, or part-aware but limited to closed-set taxonomies, which weakens zero-shot transfer. We study text-conditioned 3D part segmentation, where a free-form phrase selects a functional part on point cloud. We introduce UniPart, a feed-forward cross-modal 3D Transformer that conditions CLIP text embedding. To scale supervision, we build LangPart-1M with 160K+ Objaverse assets and 8M text to part pairs using multi-view consistent part generation. We further manually label a high-quality subset, LangPart-4K, for fine-tuning and evaluation. UniPart achieves strong zero-shot results on open-vocabulary part benchmarks and transfers to language-conditioned part grasping in real world.",
    "history": "Submission history From: Xinqiang Yu [ view email ] [v1] Fri, 11 Sep 2026 14:25:17 UTC (19,592 KB)",
    "meta": {
      "citation_title": "UniPart: Towards Zero-shot Language-Grounded 3D Part Segmentation for Embodied Interaction",
      "citation_author": "Wang, He",
      "citation_date": "2026/09/11",
      "citation_online_date": "2026/09/11",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.12898",
      "citation_arxiv_id": "2609.12898",
      "citation_abstract": "Fine-grained robotic manipulation depends on understanding parts, not only whole objects. Existing 3D foundation models tend to be either generalized but object-aware, or part-aware but limited to closed-set taxonomies, which weakens zero-shot transfer. We study text-conditioned 3D part segmentation, where a free-form phrase selects a functional part on point cloud. We introduce UniPart, a feed-forward cross-modal 3D Transformer that conditions CLIP text embedding. To scale supervision, we build LangPart-1M with 160K+ Objaverse assets and 8M text to part pairs using multi-view consistent part generation. We further manually label a high-quality subset, LangPart-4K, for fine-tuning and evaluation. UniPart achieves strong zero-shot results on open-vocabulary part benchmarks and transfers to language-conditioned part grasping in real world."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-11T22:25:17+08:00"
  },
  {
    "id": "2609.13779v1",
    "title": "Force-Aware Reinforcement Learning with Hybrid Sensorless Force Estimation for Wheeled-Legged Loco-Manipulation",
    "abstract": "Force-controlled loco-manipulation requires a whole-body policy to coordinate locomotion and arm motion while regulating end-effector interaction forces. This is challenging under floating-base dynamics and changing support contacts, particularly when end-effector force/torque sensing is unavailable for control. This paper presents a force-aware reinforcement learning approach with hybrid sensorless force estimation for wheeled-legged loco-manipulation. The proposed method provides a structured estimate of the end-effector force as an explicit policy observation, enabling force-guided contact behavior without using an end-effector force/torque sensor for control. The force estimate is obtained by combining generalized momentum observation, contact-constrained wrench projection, and temporal residual learning: the model-based components extract the physically structured part of the whole-body disturbance, while the residual network compensates the remaining motion-dependent bias. The estimated force is integrated into a mode-conditioned whole-body policy with an axis-wise force/position selector, allowing free-space motion, pure force regulation, and hybrid force/position control within one controller. Simulation results demonstrate improved sensorless force estimation and force-control performance. Hardware experiments further validate the proposed controller through quantitative valve-rotation and hybrid wiping evaluations, together with force-guided door opening and zero-force human-guided motion on a real wheeled-legged platform.",
    "history": "Submission history From: Xuanqi Zeng [ view email ] [v1] Sat, 12 Sep 2026 07:46:53 UTC (17,870 KB)",
    "meta": {
      "citation_title": "Force-Aware Reinforcement Learning with Hybrid Sensorless Force Estimation for Wheeled-Legged Loco-Manipulation",
      "citation_author": "Liu, Yun-Hui",
      "citation_date": "2026/09/12",
      "citation_online_date": "2026/09/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13779",
      "citation_arxiv_id": "2609.13779",
      "citation_abstract": "Force-controlled loco-manipulation requires a whole-body policy to coordinate locomotion and arm motion while regulating end-effector interaction forces. This is challenging under floating-base dynamics and changing support contacts, particularly when end-effector force/torque sensing is unavailable for control. This paper presents a force-aware reinforcement learning approach with hybrid sensorless force estimation for wheeled-legged loco-manipulation. The proposed method provides a structured estimate of the end-effector force as an explicit policy observation, enabling force-guided contact behavior without using an end-effector force/torque sensor for control. The force estimate is obtained by combining generalized momentum observation, contact-constrained wrench projection, and temporal residual learning: the model-based components extract the physically structured part of the whole-body disturbance, while the residual network compensates the remaining motion-dependent bias. The estimated force is integrated into a mode-conditioned whole-body policy with an axis-wise force/position selector, allowing free-space motion, pure force regulation, and hybrid force/position control within one controller. Simulation results demonstrate improved sensorless force estimation and force-control performance. Hardware experiments further validate the proposed controller through quantitative valve-rotation and hybrid wiping evaluations, together with force-guided door opening and zero-force human-guided motion on a real wheeled-legged platform."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T15:46:53+08:00"
  },
  {
    "id": "2609.13761v1",
    "title": "Learning In-Hand Object Reaching to General 6D Poses",
    "abstract": "In-hand manipulation allows multi-fingered dexterous hands to reconfigure grasped objects without releasing and regrasping them. This improves manipulation efficiency by reducing repeated grasp acquisition and large arm motions. However, most learning-based methods focus on reorientation, continuous rotation, or translation, whereas many tasks require joint control of object position and orientation. We formulate this capability as in-hand 6D object pose reaching: starting from an existing grasp, coordinated finger motions move the object to a palm-relative target pose. We present POISE (Palm-relative Object reaching In SE(3)), a sim-to-real reinforcement learning framework for this task. POISE combines diverse stable-grasp initialization, goal- and geometry-conditioned control, an adaptive 6D goal curriculum, and a compact reward scheme for pose reaching and grasp preservation. In simulation, diverse initialization raises held-out-grasp success from 40.1% to 51.5% and post-drop recovery from 33.8% to 72.9%; the curriculum raises full-range success from 6.2% to 59.5%. On hardware, the grasp-maintenance reward improves three-target sequence success from 20% to 80%. In real-world experiments, POISE reaches successive 6D targets without manual reset across multiple object geometries and wrist orientations, and recovers from external disturbances. To support further research in dexterous manipulation, we will release our code at this https URL .",
    "history": "Submission history From: Junxiao Lin [ view email ] [v1] Sat, 12 Sep 2026 07:11:37 UTC (3,859 KB)",
    "meta": {
      "citation_title": "Learning In-Hand Object Reaching to General 6D Poses",
      "citation_author": "Zhi, Weiming",
      "citation_date": "2026/09/12",
      "citation_online_date": "2026/09/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13761",
      "citation_arxiv_id": "2609.13761",
      "citation_abstract": "In-hand manipulation allows multi-fingered dexterous hands to reconfigure grasped objects without releasing and regrasping them. This improves manipulation efficiency by reducing repeated grasp acquisition and large arm motions. However, most learning-based methods focus on reorientation, continuous rotation, or translation, whereas many tasks require joint control of object position and orientation. We formulate this capability as in-hand 6D object pose reaching: starting from an existing grasp, coordinated finger motions move the object to a palm-relative target pose. We present POISE (Palm-relative Object reaching In SE(3)), a sim-to-real reinforcement learning framework for this task. POISE combines diverse stable-grasp initialization, goal- and geometry-conditioned control, an adaptive 6D goal curriculum, and a compact reward scheme for pose reaching and grasp preservation. In simulation, diverse initialization raises held-out-grasp success from 40.1% to 51.5% and post-drop recovery from 33.8% to 72.9%; the curriculum raises full-range success from 6.2% to 59.5%. On hardware, the grasp-maintenance reward improves three-target sequence success from 20% to 80%. In real-world experiments, POISE reaches successive 6D targets without manual reset across multiple object geometries and wrist orientations, and recovers from external disturbances. To support further research in dexterous manipulation, we will release our code at https://junxiaolin.github.io/poise-website/."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T15:11:37+08:00"
  },
  {
    "id": "2609.13516v1",
    "title": "Constraint-Grounded Reinforcement Learning for Variable Impedance Control in Contact-Rich Robotic Insertion",
    "abstract": "In robotic insertion under uncertain contact, the axial force limit and the appropriate controller gain vary across tasks. As a result, a single fixed gain is unlikely to remain suitable across different task conditions, making conventional impedance controllers reliant on manual retuning. To eliminate manual retuning, we propose Constraint-Grounded Reinforcement Learning (CG-RL), a variable impedance framework for online gain adaptation. Conditioned on the force limit and contact feedback, the policy outputs a residual motion, an insertion rate, and a requested gain. The controller projects this gain into the admissible range without exposing the range itself to the policy. This separation allows a single policy to operate under different force limits without retraining or manual retuning. We evaluate CG-RL on simulated oblique insertion across five training seeds. CG-RL achieves an $85.8\\pm7.7\\%$ (mean $\\pm$ SD) success rate of insertions without violating the force limit, while keeping the applied gain within the admissible range. As a comparison, a fixed-gain baseline using the midpoint gain achieves a success rate of $50.1\\%$. The policy adapts its insertion rate continuously to the specified force limit and further generalizes to more permissive force limits above the training range. In contrast, the same actor without force-limit input does not exhibit this adaptation. The applied gain is guaranteed to remain within the admissible range, while force-limit satisfaction is validated empirically rather than guaranteed formally.",
    "history": "Submission history From: Lin He [ view email ] [v1] Fri, 11 Sep 2026 20:36:38 UTC (2,138 KB)",
    "meta": {
      "citation_title": "Constraint-Grounded Reinforcement Learning for Variable Impedance Control in Contact-Rich Robotic Insertion",
      "citation_author": "Deng, Min",
      "citation_date": "2026/09/11",
      "citation_online_date": "2026/09/11",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.13516",
      "citation_arxiv_id": "2609.13516",
      "citation_abstract": "In robotic insertion under uncertain contact, the axial force limit and the appropriate controller gain vary across tasks. As a result, a single fixed gain is unlikely to remain suitable across different task conditions, making conventional impedance controllers reliant on manual retuning. To eliminate manual retuning, we propose Constraint-Grounded Reinforcement Learning (CG-RL), a variable impedance framework for online gain adaptation. Conditioned on the force limit and contact feedback, the policy outputs a residual motion, an insertion rate, and a requested gain. The controller projects this gain into the admissible range without exposing the range itself to the policy. This separation allows a single policy to operate under different force limits without retraining or manual retuning. We evaluate CG-RL on simulated oblique insertion across five training seeds. CG-RL achieves an $85.8\\pm7.7\\%$ (mean $\\pm$ SD) success rate of insertions without violating the force limit, while keeping the applied gain within the admissible range. As a comparison, a fixed-gain baseline using the midpoint gain achieves a success rate of $50.1\\%$. The policy adapts its insertion rate continuously to the specified force limit and further generalizes to more permissive force limits above the training range. In contrast, the same actor without force-limit input does not exhibit this adaptation. The applied gain is guaranteed to remain within the admissible range, while force-limit satisfaction is validated empirically rather than guaranteed formally."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-12T04:36:38+08:00"
  },
  {
    "id": "2609.14146v1",
    "title": "When Faster VLA Deployment Changes Closed-Loop Behavior: Task Success-Latency Analysis of SmolVLA Across PyTorch and ONNX Variants",
    "abstract": "Vision-language-action (VLA) deployment can reduce inference latency while changing closed-loop task behavior. We evaluate HuggingFaceVLA/smolvla_libero on an RTX 2060 (6 GB) in LIBERO Spatial and Object (MuJoCo 3.3.2, LeRobot 0.6.1, seed 42), comparing PyTorch+AMP with ONNX Runtime CUDA Execution Provider (CUDA EP). The main evaluation uses 100 episodes/suite; a paired rollout uses 300 episodes/suite. PyTorch+AMP reaches 70.0%/88.0% Spatial/Object success at 1181 ms p99. Requested-FP16 and requested-INT8 ONNX reduce tether-inspect p99 to 601 ms and 532 ms, while Spatial success falls to 41.0% and 40.0% and Object remains at 89.0%. A graph audit shows those artifacts are byte-identical FP32 graphs, so the requested-INT8 row is not operator-level INT8 quantization. A static language-width ablation (16/24/32 tokens) yields Spatial success of 41.0%, 75.0%, and 71.0%; widths 24 and 32 recover much of the Spatial drop while Object success and uniform-bench latency stay approximately stable. Width-24 ONNX Spatial success is comparable to the PyTorch+AMP baseline at roughly half the latency (Wilson intervals overlap; two-proportion chi-squared p=0.53). Context width is an important contributor in this stack; it does not account for every PyTorch-vs-ONNX difference. Deployment evaluation should jointly report latency, artifact inspection, interface constraints, and closed-loop success. Code: this https URL .",
    "history": "Submission history From: Rafiqul Islam [ view email ] [v1] Sat, 12 Sep 2026 20:56:31 UTC (30 KB) [v2] Wed, 16 Sep 2026 19:41:07 UTC (30 KB)",
    "meta": {
      "citation_title": "When Faster VLA Deployment Changes Closed-Loop Behavior: Task Success-Latency Analysis of SmolVLA Across PyTorch and ONNX Variants",
      "citation_author": "Islam, Rafiqul",
      "citation_date": "2026/09/12",
      "citation_online_date": "2026/09/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.14146",
      "citation_arxiv_id": "2609.14146",
      "citation_abstract": "Vision-language-action (VLA) deployment can reduce inference latency while changing closed-loop task behavior. We evaluate HuggingFaceVLA/smolvla_libero on an RTX 2060 (6 GB) in LIBERO Spatial and Object (MuJoCo 3.3.2, LeRobot 0.6.1, seed 42), comparing PyTorch+AMP with ONNX Runtime CUDA Execution Provider (CUDA EP). The main evaluation uses 100 episodes/suite; a paired rollout uses 300 episodes/suite. PyTorch+AMP reaches 70.0%/88.0% Spatial/Object success at 1181 ms p99. Requested-FP16 and requested-INT8 ONNX reduce tether-inspect p99 to 601 ms and 532 ms, while Spatial success falls to 41.0% and 40.0% and Object remains at 89.0%. A graph audit shows those artifacts are byte-identical FP32 graphs, so the requested-INT8 row is not operator-level INT8 quantization. A static language-width ablation (16/24/32 tokens) yields Spatial success of 41.0%, 75.0%, and 71.0%; widths 24 and 32 recover much of the Spatial drop while Object success and uniform-bench latency stay approximately stable. Width-24 ONNX Spatial success is comparable to the PyTorch+AMP baseline at roughly half the latency (Wilson intervals overlap; two-proportion chi-squared p=0.53). Context width is an important contributor in this stack; it does not account for every PyTorch-vs-ONNX difference. Deployment evaluation should jointly report latency, artifact inspection, interface constraints, and closed-loop success. Code: https://github.com/rafiqul713/smolvla-libero-onnx."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-13T04:56:31+08:00"
  },
  {
    "id": "2609.12549v1",
    "title": "STAR: Sparse Tactile Representation Learning in Vision-Tactile-Language-Action Models for Dexterous Manipulation",
    "abstract": "Dexterous manipulation requires coordinated multi-finger control and effective tactile feedback, yet learning these capabilities remains challenging due to the lack of large-scale real-world data and the difficulty of extracting effective representations from sparse tactile signals. We build a robot platform and teleoperation system to collect a 200-hour bimanual dexterous manipulation dataset with synchronized visual, tactile, and language annotations, comprising 10,576 trajectories across 65 tasks, 69.5% of which involve dexterous multi-finger manipulation. We further propose STAR, an integrated training recipe for vision-tactile-language-action (VTLA) models that addresses the spatial, temporal, and informational sparsity of tactile signals through visual-tactile joint pre-training, sparse-global tactile token representation, and sparse future tactile prediction. Trained on this dataset, STAR achieves a 61% average success rate across four real-world tasks with 100 post-training trajectories per task, demonstrating dexterous performance under task-specific post-training.",
    "history": "Submission history From: Xiangcheng Liu [ view email ] [v1] Fri, 11 Sep 2026 07:57:24 UTC (2,412 KB) [v2] Fri, 18 Sep 2026 16:43:17 UTC (2,412 KB)",
    "meta": {
      "citation_title": "STAR: Sparse Tactile Representation Learning in Vision-Tactile-Language-Action Models for Dexterous Manipulation",
      "citation_author": "Luo, Jianlan",
      "citation_date": "2026/09/11",
      "citation_online_date": "2026/09/11",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.12549",
      "citation_arxiv_id": "2609.12549",
      "citation_abstract": "Dexterous manipulation requires coordinated multi-finger control and effective tactile feedback, yet learning these capabilities remains challenging due to the lack of large-scale real-world data and the difficulty of extracting effective representations from sparse tactile signals. We build a robot platform and teleoperation system to collect a 200-hour bimanual dexterous manipulation dataset with synchronized visual, tactile, and language annotations, comprising 10,576 trajectories across 65 tasks, 69.5% of which involve dexterous multi-finger manipulation. We further propose STAR, an integrated training recipe for vision-tactile-language-action (VTLA) models that addresses the spatial, temporal, and informational sparsity of tactile signals through visual-tactile joint pre-training, sparse-global tactile token representation, and sparse future tactile prediction. Trained on this dataset, STAR achieves a 61% average success rate across four real-world tasks with 100 post-training trajectories per task, demonstrating dexterous performance under task-specific post-training."
    },
    "groups": [
      "w1"
    ],
    "event_at": "2026-09-11T15:57:24+08:00"
  },
  {
    "id": "2609.22888v1",
    "title": "Stable and Efficient Real-World Online VLA Post-Training via Asynchronous Replay-Anchored Policy Improvement",
    "abstract": "Online post-training of vision-language-action (VLA) models requires efficient use of robot interaction and reliable policy improvement from continually collected experience. We propose asynchronous Replay-Anchored Policy improvement (RAPolicy), a framework that performs rollout and learning concurrently while grounding both critic and actor updates in replayed behavior. The critic learns chunk-level values from recorded actions and constructs Bellman targets without predicting next actions, reducing computation and dependence on action-value estimates outside replay coverage. The one-step flow actor reuses the initial noise stored during rollout and learns through advantage-weighted conditional likelihood, directly supervising the action mapping used for execution. We evaluate RAPolicy across four single-task settings and one joint five-task setting in the real world, with online training budgets of only 1--2 hours. Starting from policies fine-tuned on just 10 demonstrations per task, RAPolicy rapidly adapts to new single tasks and achieves an average 86.3% success rate. In the joint multi-task setting, RAPolicy improves overall success rate from 52% to 88% while preserving performance on already reliable tasks and improving weaker capabilities. Overall, RAPolicy substantially outperforms the baselines in aggregate task success while requiring fewer human interventions, demonstrating stable policy improvement and high online training efficiency. Project page: this https URL .",
    "history": "Submission history From: Jiarui Yang [ view email ] [v1] Sat, 19 Sep 2026 08:45:18 UTC (2,723 KB)",
    "meta": {
      "citation_title": "Stable and Efficient Real-World Online VLA Post-Training via Asynchronous Replay-Anchored Policy Improvement",
      "citation_author": "Jiang, Yu-Gang",
      "citation_date": "2026/09/19",
      "citation_online_date": "2026/09/19",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.22888",
      "citation_arxiv_id": "2609.22888",
      "citation_abstract": "Online post-training of vision-language-action (VLA) models requires efficient use of robot interaction and reliable policy improvement from continually collected experience. We propose asynchronous Replay-Anchored Policy improvement (RAPolicy), a framework that performs rollout and learning concurrently while grounding both critic and actor updates in replayed behavior. The critic learns chunk-level values from recorded actions and constructs Bellman targets without predicting next actions, reducing computation and dependence on action-value estimates outside replay coverage. The one-step flow actor reuses the initial noise stored during rollout and learns through advantage-weighted conditional likelihood, directly supervising the action mapping used for execution. We evaluate RAPolicy across four single-task settings and one joint five-task setting in the real world, with online training budgets of only 1--2 hours. Starting from policies fine-tuned on just 10 demonstrations per task, RAPolicy rapidly adapts to new single tasks and achieves an average 86.3% success rate. In the joint multi-task setting, RAPolicy improves overall success rate from 52% to 88% while preserving performance on already reliable tasks and improving weaker capabilities. Overall, RAPolicy substantially outperforms the baselines in aggregate task success while requiring fewer human interventions, demonstrating stable policy improvement and high online training efficiency. Project page: https://flyfaerss.github.io/RAPolicy."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-19T16:45:18+08:00"
  },
  {
    "id": "2609.22840v1",
    "title": "ForceRFT: Refining VLA Actions through Force-Guided Residual Reinforcement Learning",
    "abstract": "Force-conditioned vision-language-action (VLA) policies can respond to contact, but when trained solely on demonstrations, their recovery behavior may be limited by demonstration coverage, and they do not learn from deployment outcomes. Human corrective imitation provides additional recovery examples, but its objective matches local action targets without explicitly optimizing task return. We present ForceRFT, a force-guided residual reinforcement learning framework that learns contact-dependent corrections from human supervision and autonomous task outcomes. A frozen, demonstration-trained SmolVLA-based prior generates force-conditioned action chunks, while a lightweight residual actor refines individual end-effector pose commands using wrist feedback acquired during chunk execution. The decision-time wrist wrench, its temporal change, and the selected base motion condition both residual correction and value estimation. Human corrections supervise the residual actor, while verified autonomous transitions train the twin critics and support value-guided updates to the same actor. Bootstrapping is restricted to autonomous segments, preventing TD credit from crossing human-intervention boundaries. Real-robot experiments on plug insertion, ring-on-peg assembly, and whiteboard wiping show higher autonomous success rates than the evaluated demonstration-trained and residual-imitation baselines. Comparisons with residual imitation support value-guided residual optimization, while plug-insertion ablations indicate the benefit of direct execution-time wrist feedback.",
    "history": "Submission history From: Yichen Wang [ view email ] [v1] Sat, 19 Sep 2026 07:16:11 UTC (663 KB)",
    "meta": {
      "citation_title": "ForceRFT: Refining VLA Actions through Force-Guided Residual Reinforcement Learning",
      "citation_author": "Li, Xiaocong",
      "citation_date": "2026/09/19",
      "citation_online_date": "2026/09/19",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.22840",
      "citation_arxiv_id": "2609.22840",
      "citation_abstract": "Force-conditioned vision-language-action (VLA) policies can respond to contact, but when trained solely on demonstrations, their recovery behavior may be limited by demonstration coverage, and they do not learn from deployment outcomes. Human corrective imitation provides additional recovery examples, but its objective matches local action targets without explicitly optimizing task return. We present ForceRFT, a force-guided residual reinforcement learning framework that learns contact-dependent corrections from human supervision and autonomous task outcomes. A frozen, demonstration-trained SmolVLA-based prior generates force-conditioned action chunks, while a lightweight residual actor refines individual end-effector pose commands using wrist feedback acquired during chunk execution. The decision-time wrist wrench, its temporal change, and the selected base motion condition both residual correction and value estimation. Human corrections supervise the residual actor, while verified autonomous transitions train the twin critics and support value-guided updates to the same actor. Bootstrapping is restricted to autonomous segments, preventing TD credit from crossing human-intervention boundaries. Real-robot experiments on plug insertion, ring-on-peg assembly, and whiteboard wiping show higher autonomous success rates than the evaluated demonstration-trained and residual-imitation baselines. Comparisons with residual imitation support value-guided residual optimization, while plug-insertion ablations indicate the benefit of direct execution-time wrist feedback."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-19T15:16:11+08:00"
  },
  {
    "id": "2609.22587v1",
    "title": "React When You Need To: Event-Triggered Asynchronous Inference for VLA Policies",
    "abstract": "Vision-Language-Action (VLA) models commonly predict action chunks, limiting their ability to react to environmental changes during execution. Existing asynchronous inference methods improve reactivity but typically rely on a fixed inference gap. In this paper, we propose an event-guided dynamic inference strategy that adapts the inference gap according to scene changes observed since the previous inference. Thereby, it simultaneously preserves motion consistency and prompt reactivity. Across static and dynamic real-world settings, our method consistently performs best, averaging 95% success and exceeding the strongest baseline by 55 percentage points. The code will be made publicly available upon acceptance. The project page is available at this https URL .",
    "history": "Submission history From: Yansong Wu [ view email ] [v1] Fri, 18 Sep 2026 21:11:23 UTC (4,021 KB)",
    "meta": {
      "citation_title": "React When You Need To: Event-Triggered Asynchronous Inference for VLA Policies",
      "citation_author": "Knoll, Alois",
      "citation_date": "2026/09/18",
      "citation_online_date": "2026/09/18",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.22587",
      "citation_arxiv_id": "2609.22587",
      "citation_abstract": "Vision-Language-Action (VLA) models commonly predict action chunks, limiting their ability to react to environmental changes during execution. Existing asynchronous inference methods improve reactivity but typically rely on a fixed inference gap. In this paper, we propose an event-guided dynamic inference strategy that adapts the inference gap according to scene changes observed since the previous inference. Thereby, it simultaneously preserves motion consistency and prompt reactivity. Across static and dynamic real-world settings, our method consistently performs best, averaging 95% success and exceeding the strongest baseline by 55 percentage points. The code will be made publicly available upon acceptance. The project page is available at https://react-when-you-need-to.github.io/."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-19T05:11:23+08:00"
  },
  {
    "id": "2609.22591v1",
    "title": "REBOOT: From Failure to Recovery - A Dataset and Benchmark for Precision Assembly",
    "abstract": "Robot learning policies fail in characteristic ways: they stall in uncertain states, drift during contact-rich alignment, and miss targets by millimetres in precision tasks. Yet training datasets consist largely of successful demonstrations, while real-world benchmarks often reduce performance to binary success. This limits both supervision for recovery and analysis of where failures occur. We introduce REBOOT (Recovery Episode Benchmark for Off-nominal Trajectories), the first robot manipulation benchmark designed around failure as a first-class signal. REBOOT contains 2,160 demonstrations across 18 precision assembly tasks, each decomposed into five shared phases: Align(pick), Engage(pick), Transport, Align(place), and Engage(place), enabling phase-level evaluation beyond terminal success. Failures are introduced across phases and paired with expert recovery trajectories that return the system to a valid continuation state. Tasks are annotated with rotational symmetry, engagement-clearance precision tier, and assembly direction through matched install-remove pairs. Failure episodes are labeled by phase and categorical failure mode, enabling attribution to kinematic stage and tolerance violation. Data includes synchronized RGB-D observations from four viewpoints and grounded natural-language descriptions of phase-level success and failure conditions. Half the dataset contains expert demonstrations; the other half contains recovery demonstrations sampled to reflect failures observed in imitation-learned policy rollouts. We benchmark action-chunked transformer, diffusion, and $\\pi_0$-FAST policies using phase-level completion rates, revealing model-specific failure points hidden by binary evaluation. Dataset and code: this https URL",
    "history": "Submission history From: Nana Yaw Owusu Ofori-Ampofo Mr [ view email ] [v1] Fri, 18 Sep 2026 21:14:54 UTC (20,761 KB)",
    "meta": {
      "citation_title": "REBOOT: From Failure to Recovery - A Dataset and Benchmark for Precision Assembly",
      "citation_author": "Thekinen, Joseph",
      "citation_date": "2026/09/18",
      "citation_online_date": "2026/09/18",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.22591",
      "citation_arxiv_id": "2609.22591",
      "citation_abstract": "Robot learning policies fail in characteristic ways: they stall in uncertain states, drift during contact-rich alignment, and miss targets by millimetres in precision tasks. Yet training datasets consist largely of successful demonstrations, while real-world benchmarks often reduce performance to binary success. This limits both supervision for recovery and analysis of where failures occur. We introduce REBOOT (Recovery Episode Benchmark for Off-nominal Trajectories), the first robot manipulation benchmark designed around failure as a first-class signal. REBOOT contains 2,160 demonstrations across 18 precision assembly tasks, each decomposed into five shared phases: Align(pick), Engage(pick), Transport, Align(place), and Engage(place), enabling phase-level evaluation beyond terminal success. Failures are introduced across phases and paired with expert recovery trajectories that return the system to a valid continuation state. Tasks are annotated with rotational symmetry, engagement-clearance precision tier, and assembly direction through matched install-remove pairs. Failure episodes are labeled by phase and categorical failure mode, enabling attribution to kinematic stage and tolerance violation. Data includes synchronized RGB-D observations from four viewpoints and grounded natural-language descriptions of phase-level success and failure conditions. Half the dataset contains expert demonstrations; the other half contains recovery demonstrations sampled to reflect failures observed in imitation-learned policy rollouts. We benchmark action-chunked transformer, diffusion, and $\\pi_0$-FAST policies using phase-level completion rates, revealing model-specific failure points hidden by binary evaluation. Dataset and code: https://nanayawoa.github.io/REBOOT"
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-19T05:14:54+08:00"
  },
  {
    "id": "2609.22538v1",
    "title": "FRAMES: Failure Recovery And Monitoring of Embodied Skills for Humanoid Loco-Manipulation",
    "abstract": "Large language model (LLM) planners can decompose natural-language instructions and select reusable robot skills, but choosing the correct skill does not guarantee successful physical execution. This gap is especially important in humanoid loco-manipulation, where errors during approach, grasping, transport, or placement can invalidate the remainder of a long-horizon plan. We present FRAMES, a failure-aware supervisory framework for the Unitree G1 humanoid that operates above the CEER whole-body controller. A Planner Agent selects subtasks through parameterized mid-level skills, while a vision-language-model-based Monitor Agent evaluates each skill using temporal multi-view observations and structured robot and contact evidence. Detected failures stop the active skill and provide grounded feedback to a Recovery Agent. The framework further includes a Memory Module for reusing prior skill experience, and geometric grounding via depth and segmentation. We independently evaluate the monitoring module of the framework in MuJoCo using 100 trials comprising 50 failed and 50 successful executions across five tasks. The monitor detects 48 of 50 failures, correctly accepts 46 of 50 successful executions, and achieves 94.0% overall accuracy. These results provide initial evidence for the monitoring component, while end-to-end evaluation of the complete recovery loop remains ongoing.",
    "history": "Submission history From: Ajay Vikram Periasami [ view email ] [v1] Fri, 18 Sep 2026 19:53:21 UTC (5,964 KB)",
    "meta": {
      "citation_title": "FRAMES: Failure Recovery And Monitoring of Embodied Skills for Humanoid Loco-Manipulation",
      "citation_author": "Cheng, Xianyi",
      "citation_date": "2026/09/18",
      "citation_online_date": "2026/09/18",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.22538",
      "citation_arxiv_id": "2609.22538",
      "citation_abstract": "Large language model (LLM) planners can decompose natural-language instructions and select reusable robot skills, but choosing the correct skill does not guarantee successful physical execution. This gap is especially important in humanoid loco-manipulation, where errors during approach, grasping, transport, or placement can invalidate the remainder of a long-horizon plan. We present FRAMES, a failure-aware supervisory framework for the Unitree G1 humanoid that operates above the CEER whole-body controller. A Planner Agent selects subtasks through parameterized mid-level skills, while a vision-language-model-based Monitor Agent evaluates each skill using temporal multi-view observations and structured robot and contact evidence. Detected failures stop the active skill and provide grounded feedback to a Recovery Agent. The framework further includes a Memory Module for reusing prior skill experience, and geometric grounding via depth and segmentation. We independently evaluate the monitoring module of the framework in MuJoCo using 100 trials comprising 50 failed and 50 successful executions across five tasks. The monitor detects 48 of 50 failures, correctly accepts 46 of 50 successful executions, and achieves 94.0% overall accuracy. These results provide initial evidence for the monitoring component, while end-to-end evaluation of the complete recovery loop remains ongoing."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-19T03:53:21+08:00"
  },
  {
    "id": "2609.22606v1",
    "title": "VT-Bridge: Bridging Pretrained Foundation VLAs to VTLAs via Lightweight Residual Adaptation",
    "abstract": "Vision-Tactile-Language-Action (VTLA) models have demonstrated clear advantages over Vision-Language-Action (VLA) models in contact-rich manipulation. However, developing VTLA models is severely constrained by the massive amounts of vision-tactile data and computational resources required. To address this bottleneck, we propose VT-Bridge, a lightweight residual adaptation strategy that bridges pretrained foundation VLAs to VTLAs. Rather than training a VTLA model from scratch or modifying the original architecture of a pretrained VLA, VT-Bridge employs an identical lightweight residual-adapter architecture across VLA backbones and uses backbone-specific weights to refine actions at the robot execution frequency. This design substantially lowers the data and training barriers. Specifically, it requires up to 50 vision-tactile demonstrations per task to fine-tune a VLA backbone and train a 0.98M-parameter residual adapter. Experiments with three representative VLA backbones ($\\pi_0$, $\\pi_{0.5}$, and SmolVLA) across four contact-rich manipulation tasks further demonstrate its consistent effectiveness across VLA architectures. On average, VT-Bridge raises the task completion rate from 11.7% with task-level VLA fine-tuning alone to 62.9%. Together, these findings demonstrate the broad applicability, effectiveness, and accessibility of VT-Bridge for contact-rich manipulation. The project page is available at this https URL .",
    "history": "Submission history From: Yansong Wu [ view email ] [v1] Fri, 18 Sep 2026 21:48:33 UTC (2,972 KB)",
    "meta": {
      "citation_title": "VT-Bridge: Bridging Pretrained Foundation VLAs to VTLAs via Lightweight Residual Adaptation",
      "citation_author": "Knoll, Alois",
      "citation_date": "2026/09/18",
      "citation_online_date": "2026/09/18",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.22606",
      "citation_arxiv_id": "2609.22606",
      "citation_abstract": "Vision-Tactile-Language-Action (VTLA) models have demonstrated clear advantages over Vision-Language-Action (VLA) models in contact-rich manipulation. However, developing VTLA models is severely constrained by the massive amounts of vision-tactile data and computational resources required. To address this bottleneck, we propose VT-Bridge, a lightweight residual adaptation strategy that bridges pretrained foundation VLAs to VTLAs. Rather than training a VTLA model from scratch or modifying the original architecture of a pretrained VLA, VT-Bridge employs an identical lightweight residual-adapter architecture across VLA backbones and uses backbone-specific weights to refine actions at the robot execution frequency. This design substantially lowers the data and training barriers. Specifically, it requires up to 50 vision-tactile demonstrations per task to fine-tune a VLA backbone and train a 0.98M-parameter residual adapter. Experiments with three representative VLA backbones ($\\pi_0$, $\\pi_{0.5}$, and SmolVLA) across four contact-rich manipulation tasks further demonstrate its consistent effectiveness across VLA architectures. On average, VT-Bridge raises the task completion rate from 11.7% with task-level VLA fine-tuning alone to 62.9%. Together, these findings demonstrate the broad applicability, effectiveness, and accessibility of VT-Bridge for contact-rich manipulation. The project page is available at https://hoxnocha.github.io/vt-bridge-web/."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-19T05:48:33+08:00"
  },
  {
    "id": "2609.22829v1",
    "title": "Whole-Body UMI: Transferring UMI Manipulation Skills to Humanoid Whole-Body Manipulation via Real-Time Motion Generation",
    "abstract": "Collecting whole-body demonstrations for humanoid manipulation mostly relies on teleoperation, which is costly and hard to scale up. The Universal Manipulation Interface (UMI) provides a scalable data collection paradigm, but end-effector trajectories alone underdetermine humanoid whole-body coordination, which is insufficient for whole-body demonstration collection. Therefore, we introduce Whole-Body UMI (WB-UMI), a task-agnostic, real-time and end-effector conditioned motion generator that decouples whole-body coordination learning from task semantics learning through a shared end-effector interface. A diffusion policy learns from native UMI demonstrations, while WB-UMI learns independently from retargeted motion capture, requiring no body trackers or paired image--whole-body demonstrations during task-specific data collection. In real deployment, an asynchronous hierarchy integrates the diffusion policy, motion generator, and a whole-body controller with latency compensation and measured-state feedback. Real-robot experiments on G1 support real-time closed-loop transfer across four tasks, achieving 90% success in drawer closing, 80% in shelf pick-and-place, 30% in ball toss, and 40% in Loco-PnP, which shows the effectiveness of this hierarchy in transferring native UMI skills to humanoid whole-body manipulation.",
    "history": "Submission history From: Yuxuan Nai [ view email ] [v1] Sat, 19 Sep 2026 07:04:01 UTC (5,359 KB) [v2] Tue, 22 Sep 2026 09:15:49 UTC (5,359 KB)",
    "meta": {
      "citation_title": "Whole-Body UMI: Transferring UMI Manipulation Skills to Humanoid Whole-Body Manipulation via Real-Time Motion Generation",
      "citation_author": "Li, Zhongyu",
      "citation_date": "2026/09/19",
      "citation_online_date": "2026/09/19",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.22829",
      "citation_arxiv_id": "2609.22829",
      "citation_abstract": "Collecting whole-body demonstrations for humanoid manipulation mostly relies on teleoperation, which is costly and hard to scale up. The Universal Manipulation Interface (UMI) provides a scalable data collection paradigm, but end-effector trajectories alone underdetermine humanoid whole-body coordination, which is insufficient for whole-body demonstration collection. Therefore, we introduce Whole-Body UMI (WB-UMI), a task-agnostic, real-time and end-effector conditioned motion generator that decouples whole-body coordination learning from task semantics learning through a shared end-effector interface. A diffusion policy learns from native UMI demonstrations, while WB-UMI learns independently from retargeted motion capture, requiring no body trackers or paired image--whole-body demonstrations during task-specific data collection. In real deployment, an asynchronous hierarchy integrates the diffusion policy, motion generator, and a whole-body controller with latency compensation and measured-state feedback. Real-robot experiments on G1 support real-time closed-loop transfer across four tasks, achieving 90% success in drawer closing, 80% in shelf pick-and-place, 30% in ball toss, and 40% in Loco-PnP, which shows the effectiveness of this hierarchy in transferring native UMI skills to humanoid whole-body manipulation."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-19T15:04:01+08:00"
  },
  {
    "id": "2609.23252v1",
    "title": "Robot World Models Are Not Invariant to How the Actions Are Written",
    "abstract": "A robot policy is trained with one of two action parameterizations: absolute joint targets, or deltas relative to the current state. The choice is a live engineering decision in robot learning, and a world model conditioned on actions inherits it silently. We show the inheritance is catastrophic. A latent dynamics model trained on one parameterization and handed the identical commanded trajectory written in the other collapses: retrieval degrades by 2.6-13.4x across three robot datasets and two morphologies, goal-conditioned action selection falls from 53% to 15%, and on PushT the two beliefs about the same future are near-orthogonal (cos = 0.067, worst case -0.377), so the predictor does not degrade gracefully, it answers a different question. This is not a distribution-shift artifact in the usual sense: the two encodings are mutually reconstructible at R^2 = 0.996 given the joint input, so no information is lost, and we give the test that separates a valid re-parameterization from a lossy summary or a sensor swap. The test rejected three of the four axes we proposed. The defect lives in the action channel, which the invariance literature for visual models does not examine: work there concerns crops, jitter and camera pose, while the parameterization of the commands goes unaudited. The repair is averaging over the two encodings, and where it goes matters. Averaging the objective restores task performance by itself; averaging the outputs, safe for probabilities by concavity, is not available for direction-valued prediction, where the normalized mean can score below every member of the orbit. What objective-averaging leaves behind is the tail: worst-case agreement stays at 0.78, a disagreement penalty closes it to 0.995, and over a latent rollout it is the difference between a worst case that erodes and one that holds. On PushT, averaging alone does not repair the axis.",
    "history": "Submission history From: Ahmed Karim [ view email ] [v1] Sat, 19 Sep 2026 23:41:03 UTC (454 KB)",
    "meta": {
      "citation_title": "Robot World Models Are Not Invariant to How the Actions Are Written",
      "citation_author": "Chlon, Leon",
      "citation_date": "2026/09/19",
      "citation_online_date": "2026/09/19",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.23252",
      "citation_arxiv_id": "2609.23252",
      "citation_abstract": "A robot policy is trained with one of two action parameterizations: absolute joint targets, or deltas relative to the current state. The choice is a live engineering decision in robot learning, and a world model conditioned on actions inherits it silently. We show the inheritance is catastrophic. A latent dynamics model trained on one parameterization and handed the identical commanded trajectory written in the other collapses: retrieval degrades by 2.6-13.4x across three robot datasets and two morphologies, goal-conditioned action selection falls from 53% to 15%, and on PushT the two beliefs about the same future are near-orthogonal (cos = 0.067, worst case -0.377), so the predictor does not degrade gracefully, it answers a different question. This is not a distribution-shift artifact in the usual sense: the two encodings are mutually reconstructible at R^2 = 0.996 given the joint input, so no information is lost, and we give the test that separates a valid re-parameterization from a lossy summary or a sensor swap. The test rejected three of the four axes we proposed. The defect lives in the action channel, which the invariance literature for visual models does not examine: work there concerns crops, jitter and camera pose, while the parameterization of the commands goes unaudited. The repair is averaging over the two encodings, and where it goes matters. Averaging the objective restores task performance by itself; averaging the outputs, safe for probabilities by concavity, is not available for direction-valued prediction, where the normalized mean can score below every member of the orbit. What objective-averaging leaves behind is the tail: worst-case agreement stays at 0.78, a disagreement penalty closes it to 0.995, and over a latent rollout it is the difference between a worst case that erodes and one that holds. On PushT, averaging alone does not repair the axis."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-20T07:41:03+08:00"
  },
  {
    "id": "2609.23800v1",
    "title": "ContactDP: Contact-Guided Diffusion Policy for Tight Insertion Tasks",
    "abstract": "High-precision connector insertion remains challenging for robotic systems due to tight mechanical tolerances, partial observability during contact, and multimodal uncertainty arising from occlusion and contact ambiguity. Successful insertion requires closed-loop contact guidance that continuously integrates global alignment cues with local contact feedback to produce stable corrective actions under interaction. In this work, we present ContactDP (Contact-Guided Diffusion Policy for Tight Insertion Tasks), a multimodal diffusion-policy framework for contact-rich insertion. ContactDP jointly integrates wrist RGB observations, fingertip tactile sensing, and wrist-mounted force-torque measurements to infer contact state and generate temporally consistent corrective motions during insertion. To ensure stable execution under contact, the learned policy operates together with a hybrid position-force controller that provides compliant low-level interaction. We evaluate our approach on a suite of industrial-grade connector insertion tasks with varying connector geometries, grasp conditions, and initial misalignment. Across all tasks, ContactDP significantly outperforms vision-only diffusion policies for performance, reliability and generalization.",
    "history": "Submission history From: Devesh Jha [ view email ] [v1] Sun, 20 Sep 2026 18:40:48 UTC (6,036 KB)",
    "meta": {
      "citation_title": "ContactDP: Contact-Guided Diffusion Policy for Tight Insertion Tasks",
      "citation_author": "Jha, Devesh K.",
      "citation_date": "2026/09/20",
      "citation_online_date": "2026/09/20",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.23800",
      "citation_arxiv_id": "2609.23800",
      "citation_abstract": "High-precision connector insertion remains challenging for robotic systems due to tight mechanical tolerances, partial observability during contact, and multimodal uncertainty arising from occlusion and contact ambiguity. Successful insertion requires closed-loop contact guidance that continuously integrates global alignment cues with local contact feedback to produce stable corrective actions under interaction. In this work, we present ContactDP (Contact-Guided Diffusion Policy for Tight Insertion Tasks), a multimodal diffusion-policy framework for contact-rich insertion. ContactDP jointly integrates wrist RGB observations, fingertip tactile sensing, and wrist-mounted force-torque measurements to infer contact state and generate temporally consistent corrective motions during insertion. To ensure stable execution under contact, the learned policy operates together with a hybrid position-force controller that provides compliant low-level interaction. We evaluate our approach on a suite of industrial-grade connector insertion tasks with varying connector geometries, grasp conditions, and initial misalignment. Across all tasks, ContactDP significantly outperforms vision-only diffusion policies for performance, reliability and generalization."
    },
    "groups": [],
    "event_at": "2026-09-21T02:40:48+08:00",
    "exclusion": "UTC date falls in the following Beijing week; not selected for the closed week"
  },
  {
    "id": "2609.23666v1",
    "title": "UniPoint: Unified Point-Level Sensor Fusion for Humanoid Locomotion Across Challenging Terrains",
    "abstract": "Open-world deployment requires humanoid robots to cross highly heterogeneous terrain safely, with perception that simultaneously provides wide coverage, local accuracy, and redundancy against sensor failure. Existing approaches struggle to satisfy all three: one forward depth camera or nearby height sampling covers too little; odometry-corrected elevation maps drift under aggressive motion and miss thin vertical structures; image-level encoding costs grow with camera count. We present UniPoint, a humanoid whole-body locomotion framework built on multi-source point-level sensor fusion. Measurements from a 360° light detection and ranging (LiDAR) sensor and two depth cameras are early-fused into one base-frame point set. Voxelization resamples it to a fixed number of tokens encoded by linear self-attention and proprioception-queried cross-attention, decoupling forward cost from sensor count. The point set retains standing thin barriers; a single-modality failure removes only part of the tokens, so the policy degrades gracefully. A single training run with terrain-aware rewards, perception-degradation injection, and domain randomization produces one policy for all eight terrain types, deployed on an onboard RK3588 without fine-tuning. On a DR02 humanoid, 20 trials at each of nine real-world settings over seven terrain types validate the policy on 70-cm-high platforms, 100-cm gaps, thin barriers, and sparse or narrow footholds; it also generalizes zero-shot outdoors.",
    "history": "Submission history From: Sicen Li [ view email ] [v1] Sun, 20 Sep 2026 14:23:30 UTC (7,223 KB)",
    "meta": {
      "citation_title": "UniPoint: Unified Point-Level Sensor Fusion for Humanoid Locomotion Across Challenging Terrains",
      "citation_author": "Wu, Jun",
      "citation_date": "2026/09/20",
      "citation_online_date": "2026/09/20",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.23666",
      "citation_arxiv_id": "2609.23666",
      "citation_abstract": "Open-world deployment requires humanoid robots to cross highly heterogeneous terrain safely, with perception that simultaneously provides wide coverage, local accuracy, and redundancy against sensor failure. Existing approaches struggle to satisfy all three: one forward depth camera or nearby height sampling covers too little; odometry-corrected elevation maps drift under aggressive motion and miss thin vertical structures; image-level encoding costs grow with camera count. We present UniPoint, a humanoid whole-body locomotion framework built on multi-source point-level sensor fusion. Measurements from a 360{\\deg} light detection and ranging (LiDAR) sensor and two depth cameras are early-fused into one base-frame point set. Voxelization resamples it to a fixed number of tokens encoded by linear self-attention and proprioception-queried cross-attention, decoupling forward cost from sensor count. The point set retains standing thin barriers; a single-modality failure removes only part of the tokens, so the policy degrades gracefully. A single training run with terrain-aware rewards, perception-degradation injection, and domain randomization produces one policy for all eight terrain types, deployed on an onboard RK3588 without fine-tuning. On a DR02 humanoid, 20 trials at each of nine real-world settings over seven terrain types validate the policy on 70-cm-high platforms, 100-cm gaps, thin barriers, and sparse or narrow footholds; it also generalizes zero-shot outdoors."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-20T22:23:30+08:00"
  },
  {
    "id": "2609.33007v1",
    "title": "CAPEX: Efficiently Distilling Foundation Model Behavior into Deployable Robot Policies through Experience-Adaptive Reasoning",
    "abstract": "Robot learning has largely relied on human-teleoperated demonstrations to acquire effective learnable behaviors. However, human-operated data collection processes can be unintuitive, difficult to scale, and inherently asynchronous. We explore an alternative: distilling physical behavior from general-purpose multimodal foundation models into deployable robot policies by using the foundation model itself as an autonomous demonstrator. While sufficiently capable models can generate successful zero-shot manipulation trajectories, repeatedly invoking them during physical execution is slow and expensive, limiting their utility as scalable data generators. As a solution, we introduce CAPEX, an experience-conditioned demonstration collection framework that uses execution experience from previous attempts to adapt how frequently the foundation model must observe, reason, and replan. We evaluate across RoboCasa tasks and on physical Franka and bimanual YAM-arm platforms, measuring task success, model calls, token usage, collection time, and cost. We further train Diffusion Policy and ACT on matched sets of human-teleoperated and foundation-model-generated demonstrations to evaluate the downstream learning value of autonomously collected data. We find that CAPEX increases the number of successful demonstrations by 4.3x while reducing the cost per successful demonstration by 80%. Policies trained on CAPEX-generated data approach the performance of those trained on matched human demonstrations; with longer training, this gap largely closes for policies trained from scratch. These results suggest that foundation models can serve as scalable sources of reusable robot experience. Project page: this https URL",
    "history": "Submission history From: Shivam Aarya [ view email ] [v1] Sat, 26 Sep 2026 23:05:46 UTC (3,356 KB)",
    "meta": {
      "citation_title": "CAPEX: Efficiently Distilling Foundation Model Behavior into Deployable Robot Policies through Experience-Adaptive Reasoning",
      "citation_author": "Kira, Zsolt",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.33007",
      "citation_arxiv_id": "2609.33007",
      "citation_abstract": "Robot learning has largely relied on human-teleoperated demonstrations to acquire effective learnable behaviors. However, human-operated data collection processes can be unintuitive, difficult to scale, and inherently asynchronous. We explore an alternative: distilling physical behavior from general-purpose multimodal foundation models into deployable robot policies by using the foundation model itself as an autonomous demonstrator. While sufficiently capable models can generate successful zero-shot manipulation trajectories, repeatedly invoking them during physical execution is slow and expensive, limiting their utility as scalable data generators. As a solution, we introduce CAPEX, an experience-conditioned demonstration collection framework that uses execution experience from previous attempts to adapt how frequently the foundation model must observe, reason, and replan. We evaluate across RoboCasa tasks and on physical Franka and bimanual YAM-arm platforms, measuring task success, model calls, token usage, collection time, and cost. We further train Diffusion Policy and ACT on matched sets of human-teleoperated and foundation-model-generated demonstrations to evaluate the downstream learning value of autonomously collected data. We find that CAPEX increases the number of successful demonstrations by 4.3x while reducing the cost per successful demonstration by 80%. Policies trained on CAPEX-generated data approach the performance of those trained on matched human demonstrations; with longer training, this gap largely closes for policies trained from scratch. These results suggest that foundation models can serve as scalable sources of reusable robot experience. Project page: https://capex-paper.github.io/"
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-27T07:05:46+08:00"
  },
  {
    "id": "2609.32862v1",
    "title": "RoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents",
    "abstract": "A foundation model should not act in isolation as an embodied agent. Yet, existing methods often optimize individual components of the agent stack, such as memory, context, skills, or action interfaces, rather than treating the supporting system itself as a unified policy. Moreover, interaction alone does not yield self-improvement unless execution experience is converted into persistent, validated system changes. We therefore propose RoboFoundry, the first embodied agentic framework that formulates this process as Self-Evolving System-as-Policy. RoboFoundry diagnoses capability gaps in decision-making and memory management, converts execution traces into validated task-specific system updates, and promotes recurring improvements to the general system. Evolution operates over two complementary surfaces: a context system that manages active internal context and persistent file-system memory, and a hierarchical skill system that organizes atomic skills, reusable compositions, and failure-conditioned recovery. A shared semantic interface separates embodiment-invariant decisions from embodiment-specific execution, allowing evolved system capabilities to transfer across heterogeneous robots. On EmbodiedBench, RoboFoundry achieves state-of-the-art performance, notably improving GPT-5.5 by 27.8%. It also brings Qwen3.7-Plus to near parity with GPT-5.5 (70.3% vs. 72.7%), showing consistent gains from system-as-policy evolution across foundation models. For long-horizon memory, RoboFoundry outperforms all baselines on RoboMemArena by at least 39.0%, even against methods assisted by external foundation models. On LIBERO-PRO, it further outperforms Cap-Agent0 by 243.8%-679.7% across all perturbation types. In real-world deployments, RoboFoundry demonstrates zero-shot transfer and online evolution across robots and tasks, highlighting its potential for fully autonomous embodied agents.",
    "history": "Submission history From: Jingsong Liang [ view email ] [v1] Sat, 26 Sep 2026 18:41:03 UTC (18,247 KB)",
    "meta": {
      "citation_title": "RoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents",
      "citation_author": "Lv, Chen",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32862",
      "citation_arxiv_id": "2609.32862",
      "citation_abstract": "A foundation model should not act in isolation as an embodied agent. Yet, existing methods often optimize individual components of the agent stack, such as memory, context, skills, or action interfaces, rather than treating the supporting system itself as a unified policy. Moreover, interaction alone does not yield self-improvement unless execution experience is converted into persistent, validated system changes. We therefore propose RoboFoundry, the first embodied agentic framework that formulates this process as Self-Evolving System-as-Policy. RoboFoundry diagnoses capability gaps in decision-making and memory management, converts execution traces into validated task-specific system updates, and promotes recurring improvements to the general system. Evolution operates over two complementary surfaces: a context system that manages active internal context and persistent file-system memory, and a hierarchical skill system that organizes atomic skills, reusable compositions, and failure-conditioned recovery. A shared semantic interface separates embodiment-invariant decisions from embodiment-specific execution, allowing evolved system capabilities to transfer across heterogeneous robots. On EmbodiedBench, RoboFoundry achieves state-of-the-art performance, notably improving GPT-5.5 by 27.8%. It also brings Qwen3.7-Plus to near parity with GPT-5.5 (70.3% vs. 72.7%), showing consistent gains from system-as-policy evolution across foundation models. For long-horizon memory, RoboFoundry outperforms all baselines on RoboMemArena by at least 39.0%, even against methods assisted by external foundation models. On LIBERO-PRO, it further outperforms Cap-Agent0 by 243.8%-679.7% across all perturbation types. In real-world deployments, RoboFoundry demonstrates zero-shot transfer and online evolution across robots and tasks, highlighting its potential for fully autonomous embodied agents."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-27T02:41:03+08:00"
  },
  {
    "id": "2609.32801v1",
    "title": "PlanGuard: A Guardrail for Multi-Step Plan Safety in Embodied Agents",
    "abstract": "Embodied task planners may produce multi-step plans whose subtask dependencies and interactions with the environment create physical risks during execution. Yet existing safeguards overlook such compositional risks, as general-purpose guardrails focus on semantic harm and embodied safety detectors assess subtasks in isolation. To address this gap, we introduce PlanGuard, the first pre-execution detector that evaluates the physical safety of a complete multi-step plan in its current environment. For training and evaluation, we construct a Multi-Step Plan Safety (MSP-Safe) dataset through paired task construction, plan generation using diverse planners, and safety annotation by three judges. Task-oriented SFT on MSP-Safe establishes fundamental plan-safety assessment capabilities, yet a substantial gap remains between compact models suitable for real-time deployment and stronger but costlier large models. Accordingly, we propose Strong-Teacher Adaptive Compensation for On-Policy Distillation (STAC-OPD), which provides compact models with adaptive strong-teacher supervision along their on-policy trajectories. It combines token-level distribution transfer from a fine-tuned strong teacher with probability-routed sequence-level compensation, retaining student-generated targets when the student favors the reference safety decision and using teacher-reconstructed targets otherwise. Across all test subsets, PlanGuard-2B achieves average 87.15% ACC and 87.21% F1, demonstrating effective whole-plan physical-risk detection at compact model scale. Code and dataset will be publicly released.",
    "history": "Submission history From: Junchi Chen [ view email ] [v1] Sat, 26 Sep 2026 17:22:26 UTC (3,734 KB)",
    "meta": {
      "citation_title": "PlanGuard: A Guardrail for Multi-Step Plan Safety in Embodied Agents",
      "citation_author": "Yu, Nenghai",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32801",
      "citation_arxiv_id": "2609.32801",
      "citation_abstract": "Embodied task planners may produce multi-step plans whose subtask dependencies and interactions with the environment create physical risks during execution. Yet existing safeguards overlook such compositional risks, as general-purpose guardrails focus on semantic harm and embodied safety detectors assess subtasks in isolation. To address this gap, we introduce PlanGuard, the first pre-execution detector that evaluates the physical safety of a complete multi-step plan in its current environment. For training and evaluation, we construct a Multi-Step Plan Safety (MSP-Safe) dataset through paired task construction, plan generation using diverse planners, and safety annotation by three judges. Task-oriented SFT on MSP-Safe establishes fundamental plan-safety assessment capabilities, yet a substantial gap remains between compact models suitable for real-time deployment and stronger but costlier large models. Accordingly, we propose Strong-Teacher Adaptive Compensation for On-Policy Distillation (STAC-OPD), which provides compact models with adaptive strong-teacher supervision along their on-policy trajectories. It combines token-level distribution transfer from a fine-tuned strong teacher with probability-routed sequence-level compensation, retaining student-generated targets when the student favors the reference safety decision and using teacher-reconstructed targets otherwise. Across all test subsets, PlanGuard-2B achieves average 87.15% ACC and 87.21% F1, demonstrating effective whole-plan physical-risk detection at compact model scale. Code and dataset will be publicly released."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-27T01:22:26+08:00"
  },
  {
    "id": "2609.32698v1",
    "title": "SEES: A Self-Evolving Embodied System via Failure-Guided VLA Policy Adaptation",
    "abstract": "Recent vision-language-action (VLA) policies demonstrate promising generalization across diverse short-horizon tasks. However, they remain unreliable on long-horizon tasks, partly because the large-scale training data is biased toward single-stage manipulation tasks that are cheaper to demonstrate. A single weak atomic skill can cause failures across multiple multi-stage tasks. To address such failures, existing methods often require experts to identify the bottleneck and provide additional demonstrations, making the improvement costly and potentially impractical after deployment. To this end, we present a Self-Evolving Embodied System (SEES) that learns from failures and improves the VLA policy without additional expert demonstrations. SEES decomposes long-horizon tasks into atomic tasks and routes them to corresponding family policies. Each family consists of related atomic skills that share one VLA adapter. During execution, the system automatically monitors atomic-task outcomes to identify the most frequently failing atomic skills as the current bottlenecks. To overcome these bottlenecks, SEES constructs tailored RL tasks in simulation by restoring previously encountered states and generating task-specific success criteria with an LLM. Online RL updates the shared family adapters to promote positive transfer among related atomic skills and cumulative improvement across evolution rounds. Extensive experiments show that SEES can be integrated with different VLA backbones to progressively improve their long-horizon performance. We also observe continued improvement on unseen tasks, providing evidence of transfer beyond the evolution settings.",
    "history": "Submission history From: Tianyu Huang [ view email ] [v1] Sat, 26 Sep 2026 14:58:21 UTC (14,842 KB)",
    "meta": {
      "citation_title": "SEES: A Self-Evolving Embodied System via Failure-Guided VLA Policy Adaptation",
      "citation_author": "Gong, Mingming",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32698",
      "citation_arxiv_id": "2609.32698",
      "citation_abstract": "Recent vision-language-action (VLA) policies demonstrate promising generalization across diverse short-horizon tasks. However, they remain unreliable on long-horizon tasks, partly because the large-scale training data is biased toward single-stage manipulation tasks that are cheaper to demonstrate. A single weak atomic skill can cause failures across multiple multi-stage tasks. To address such failures, existing methods often require experts to identify the bottleneck and provide additional demonstrations, making the improvement costly and potentially impractical after deployment. To this end, we present a Self-Evolving Embodied System (SEES) that learns from failures and improves the VLA policy without additional expert demonstrations. SEES decomposes long-horizon tasks into atomic tasks and routes them to corresponding family policies. Each family consists of related atomic skills that share one VLA adapter. During execution, the system automatically monitors atomic-task outcomes to identify the most frequently failing atomic skills as the current bottlenecks. To overcome these bottlenecks, SEES constructs tailored RL tasks in simulation by restoring previously encountered states and generating task-specific success criteria with an LLM. Online RL updates the shared family adapters to promote positive transfer among related atomic skills and cumulative improvement across evolution rounds. Extensive experiments show that SEES can be integrated with different VLA backbones to progressively improve their long-horizon performance. We also observe continued improvement on unseen tasks, providing evidence of transfer beyond the evolution settings."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-26T22:58:21+08:00"
  },
  {
    "id": "2609.32634v1",
    "title": "PF-RL: Progress Field Reinforcement Learning via Goal-Conditioned Value Geometry for Vision-Language-Action Models",
    "abstract": "Reinforcement Fine-Tuning~(RFT) has emerged as a promising paradigm for improving Vision-Language-Action~(VLA) policies, yet sparse task-level outcomes provide limited credit for intermediate transitions, especially in long-horizon manipulation. A natural approach is to model intermediate task progress and use it as dense feedback for policy improvement. Despite their architectural differences, existing progress-aware methods commonly formulate task progress as an explicit scalar prediction, providing limited structure for modeling how intermediate observations relate to the task goal, which may hinder effective transition-level credit assignment. We introduce Progress Field Reinforcement Learning (PF-RL), which learns a structured goal-conditioned progress representation over pretrained VLA features and converts it into dense credit for policy optimization. A lightweight shared Progress Field head maps current and goal representations into a compact progress space, where geometric distance induces goal-conditioned value, while complementary temporal and goal-structure objectives shape the learned geometry. Transition-level value changes naturally yield dense progress advantages, enabling fine-grained credit assignment for both offline policy improvement and online reinforcement fine-tuning. Extensive experiments on LIBERO, RoboTwin2.0, and real-world bimanual manipulation tasks show that PF-RL consistently improves policy performance over strong supervised fine-tuning, reinforcement fine-tuning, and progress-aware baselines.",
    "history": "Submission history From: Yunpeng Qing [ view email ] [v1] Sat, 26 Sep 2026 13:53:11 UTC (14,303 KB)",
    "meta": {
      "citation_title": "PF-RL: Progress Field Reinforcement Learning via Goal-Conditioned Value Geometry for Vision-Language-Action Models",
      "citation_author": "Zou, Changqing",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32634",
      "citation_arxiv_id": "2609.32634",
      "citation_abstract": "Reinforcement Fine-Tuning~(RFT) has emerged as a promising paradigm for improving Vision-Language-Action~(VLA) policies, yet sparse task-level outcomes provide limited credit for intermediate transitions, especially in long-horizon manipulation. A natural approach is to model intermediate task progress and use it as dense feedback for policy improvement. Despite their architectural differences, existing progress-aware methods commonly formulate task progress as an explicit scalar prediction, providing limited structure for modeling how intermediate observations relate to the task goal, which may hinder effective transition-level credit assignment. We introduce Progress Field Reinforcement Learning (PF-RL), which learns a structured goal-conditioned progress representation over pretrained VLA features and converts it into dense credit for policy optimization. A lightweight shared Progress Field head maps current and goal representations into a compact progress space, where geometric distance induces goal-conditioned value, while complementary temporal and goal-structure objectives shape the learned geometry. Transition-level value changes naturally yield dense progress advantages, enabling fine-grained credit assignment for both offline policy improvement and online reinforcement fine-tuning. Extensive experiments on LIBERO, RoboTwin2.0, and real-world bimanual manipulation tasks show that PF-RL consistently improves policy performance over strong supervised fine-tuning, reinforcement fine-tuning, and progress-aware baselines."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-26T21:53:11+08:00"
  },
  {
    "id": "2609.32550v1",
    "title": "Are Vision-Language-Action Models Robust to One-Step Observation Perturbations?",
    "abstract": "Understanding the safety risks of vision-language-action (VLA) models is essential for their deployment in the physical world. Existing safety research has mainly considered persistent perturbations that are applied continuously to observations throughout an episode. However, momentary observation corruption, in which observations are severely perturbed only briefly within an episode, remains an underexplored safety threat. To address this gap, this work investigates robustness to one-step perturbations applied at a single time step per episode. Our experiments reveal that these perturbations substantially degrade VLA performance and that their impact depends on the action chunk execution length. Based on them, we propose CARE, which dynamically selects the execution length based on consistency with the previously predicted action chunk. CARE improves robustness with low computational overhead while preserving clean performance by selecting shorter execution lengths only under perturbations.",
    "history": "Submission history From: Shojiro Yamabe [ view email ] [v1] Sat, 26 Sep 2026 12:29:30 UTC (3,973 KB)",
    "meta": {
      "citation_title": "Are Vision-Language-Action Models Robust to One-Step Observation Perturbations?",
      "citation_author": "Sakuma, Jun",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32550",
      "citation_arxiv_id": "2609.32550",
      "citation_abstract": "Understanding the safety risks of vision-language-action (VLA) models is essential for their deployment in the physical world. Existing safety research has mainly considered persistent perturbations that are applied continuously to observations throughout an episode. However, momentary observation corruption, in which observations are severely perturbed only briefly within an episode, remains an underexplored safety threat. To address this gap, this work investigates robustness to one-step perturbations applied at a single time step per episode. Our experiments reveal that these perturbations substantially degrade VLA performance and that their impact depends on the action chunk execution length. Based on them, we propose CARE, which dynamically selects the execution length based on consistency with the previously predicted action chunk. CARE improves robustness with low computational overhead while preserving clean performance by selecting shorter execution lengths only under perturbations."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-26T20:29:30+08:00"
  },
  {
    "id": "2609.32313v1",
    "title": "MemTransfer: Benchmarking Memory Beyond Matched Experience in Embodied Decision-Making",
    "abstract": "Memory lets an embodied agent reuse past experience, yet retaining useful information does not ensure that the agent can apply it when conditions change. We present MemTransfer, a benchmark comparing six memory representations, a working-memory baseline and five representations of past experience, under a shared frozen vision-language-model policy. It comprises 100 navigation cases across ten task types in a simulated warehouse, with expert demonstrations supplying the history. Three comparisons vary the starting pose, route availability, and amount and task relevance of history. With one demonstration per task, Full-context and Episodic memory reach 95.3% and 100.0% success at the original demonstration start, but lose 48-49 percentage points at a new test start. Summary changes little between these two test starts, yet with four demonstrations per task it retains a smaller fraction of its unchanged-route success after blocking (39.3%) than Working memory (44.8%) or the two trajectory memories (56-58%). At the new test start, increasing from one to four relevant demonstrations raises Episodic success by 14.3 percentage points, while the other evaluated representations gain no more than 1.3 percentage points. Replacing half of the relevant histories with other-task experience lowers success for both trajectory memories. These results show that robustness to one kind of mismatch does not imply robustness to another, motivating evaluation of both stored information and its use at decision time.",
    "history": "Submission history From: Xianjie Dai [ view email ] [v1] Sat, 26 Sep 2026 07:14:16 UTC (3,229 KB)",
    "meta": {
      "citation_title": "MemTransfer: Benchmarking Memory Beyond Matched Experience in Embodied Decision-Making",
      "citation_author": "Huang, Heye",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32313",
      "citation_arxiv_id": "2609.32313",
      "citation_abstract": "Memory lets an embodied agent reuse past experience, yet retaining useful information does not ensure that the agent can apply it when conditions change. We present MemTransfer, a benchmark comparing six memory representations, a working-memory baseline and five representations of past experience, under a shared frozen vision-language-model policy. It comprises 100 navigation cases across ten task types in a simulated warehouse, with expert demonstrations supplying the history. Three comparisons vary the starting pose, route availability, and amount and task relevance of history. With one demonstration per task, Full-context and Episodic memory reach 95.3% and 100.0% success at the original demonstration start, but lose 48-49 percentage points at a new test start. Summary changes little between these two test starts, yet with four demonstrations per task it retains a smaller fraction of its unchanged-route success after blocking (39.3%) than Working memory (44.8%) or the two trajectory memories (56-58%). At the new test start, increasing from one to four relevant demonstrations raises Episodic success by 14.3 percentage points, while the other evaluated representations gain no more than 1.3 percentage points. Replacing half of the relevant histories with other-task experience lowers success for both trajectory memories. These results show that robustness to one kind of mismatch does not imply robustness to another, motivating evaluation of both stored information and its use at decision time."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-26T15:14:16+08:00"
  },
  {
    "id": "2609.32236v1",
    "title": "RoboFFT: Finetuning generative robot policy via online reinforcement learning with forward process",
    "abstract": "Generative models, such as diffusion and flow-based models, have shown strong promise for robot policy learning by capturing complex and multimodal action distributions from demonstrations. However, policies trained solely with imitation learning often suffer from imperfect demonstrations and distributional shifts, while further improvement typically requires additional expert data. Reinforcement learning offers a natural solution through environment interaction, but effectively finetuning generative robot policies remains challenging due to the intractability of likelihood estimation. In this work, we propose RoboFFT, a forward-process reinforcement learning framework for finetuning generative robot policies, which applies forward noising to sampled actions and uses the weighted score / flow matching loss to construct a surrogate policy ratio for PPO-style updates. We evaluate RoboFFT with popular generative robot policies on representative simulation benchmarks, including long-horizon planning and sparse reward settings. Extensive experiments and analysis demonstrate that RoboFFT consistently improves performance while achieving better stability and training efficiency. We further integrate RoboFFT into a real world RL framework and demonstrate its effectiveness in real world tasks. Project website: this https URL .",
    "history": "Submission history From: Yu Li [ view email ] [v1] Sat, 26 Sep 2026 04:52:58 UTC (7,373 KB)",
    "meta": {
      "citation_title": "RoboFFT: Finetuning generative robot policy via online reinforcement learning with forward process",
      "citation_author": "Yang, Yaodong",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32236",
      "citation_arxiv_id": "2609.32236",
      "citation_abstract": "Generative models, such as diffusion and flow-based models, have shown strong promise for robot policy learning by capturing complex and multimodal action distributions from demonstrations. However, policies trained solely with imitation learning often suffer from imperfect demonstrations and distributional shifts, while further improvement typically requires additional expert data. Reinforcement learning offers a natural solution through environment interaction, but effectively finetuning generative robot policies remains challenging due to the intractability of likelihood estimation. In this work, we propose RoboFFT, a forward-process reinforcement learning framework for finetuning generative robot policies, which applies forward noising to sampled actions and uses the weighted score / flow matching loss to construct a surrogate policy ratio for PPO-style updates. We evaluate RoboFFT with popular generative robot policies on representative simulation benchmarks, including long-horizon planning and sparse reward settings. Extensive experiments and analysis demonstrate that RoboFFT consistently improves performance while achieving better stability and training efficiency. We further integrate RoboFFT into a real world RL framework and demonstrate its effectiveness in real world tasks. Project website: https://student-of-holmes.github.io/RoboFFT/."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-26T12:52:58+08:00"
  },
  {
    "id": "2609.32158v1",
    "title": "FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation",
    "abstract": "Moving obstacles can block a previously clear route while a quadruped robot executes a motion command. We investigate whether predicting changing clearance improves navigation when motion selection accounts for the robot footprint and the time needed to react and brake. We present FutureRay, which predicts ranges across viewing directions and future times, together with encounter risk, from depth-derived range history and observable robot motion. Training emphasizes near-term clearance and penalizes errors that overstate available space. A local planner queries the same forecast for candidate headings and combines it with current observations to check clearance around the robot footprint. Model-based reaction--braking limits guide speed selection, and the resulting velocity commands are passed to a fixed locomotion policy. In paired evaluations on 60 static and dynamic simulation scenes, FutureRay achieves 93.3% completion, compared with 75.0% for current-range persistence and 80.0% for Cartesian Kalman rollout, with perception, planning, and locomotion held fixed. FutureRay also records fewer collisions than both baselines. Qualitative trials on a physical quadruped show avoidance initiated while an obstacle is approaching the route, followed by renewed goal progress. These results show that joint range and encounter-risk prediction can improve obstacle avoidance without retraining the locomotion policy.",
    "history": "Submission history From: Tianhao Zang [ view email ] [v1] Sat, 26 Sep 2026 02:25:49 UTC (13,915 KB)",
    "meta": {
      "citation_title": "FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation",
      "citation_author": "Zhang, Wei",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32158",
      "citation_arxiv_id": "2609.32158",
      "citation_abstract": "Moving obstacles can block a previously clear route while a quadruped robot executes a motion command. We investigate whether predicting changing clearance improves navigation when motion selection accounts for the robot footprint and the time needed to react and brake. We present FutureRay, which predicts ranges across viewing directions and future times, together with encounter risk, from depth-derived range history and observable robot motion. Training emphasizes near-term clearance and penalizes errors that overstate available space. A local planner queries the same forecast for candidate headings and combines it with current observations to check clearance around the robot footprint. Model-based reaction--braking limits guide speed selection, and the resulting velocity commands are passed to a fixed locomotion policy. In paired evaluations on 60 static and dynamic simulation scenes, FutureRay achieves 93.3% completion, compared with 75.0% for current-range persistence and 80.0% for Cartesian Kalman rollout, with perception, planning, and locomotion held fixed. FutureRay also records fewer collisions than both baselines. Qualitative trials on a physical quadruped show avoidance initiated while an obstacle is approaching the route, followed by renewed goal progress. These results show that joint range and encounter-risk prediction can improve obstacle avoidance without retraining the locomotion policy."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-26T10:25:49+08:00"
  },
  {
    "id": "2609.32767v1",
    "title": "CLAP: Closed-Loop Alignment with Pressure for Precise Suction Manipulation",
    "abstract": "Stacking and palletising demand precise placement: error left in one layer is inherited by the next, and a flat pad offers no feature to funnel a wrong pose into the right one. Top-down suction suits such dense arrangements, and suction has already been brought into vision-language-action (VLA) policies. What that work does not report, however, is a policy conditioned on a measured vacuum signal, or one that uses it to abandon an action already under way. Vision does not settle the question here, because at the moment it matters the cup and the face it holds occlude each other. We present CLAP, which makes the attachment state observable through a pressure module tapped into the vacuum line. The decoded reading replaces the suction command in the policy's proprioception, is fused with the visual features, and terminates the open-loop execution window so that the policy re-infers from a fresh observation. For data, we record goal-state disassembly on the physical robot and reverse the joint-state sequence offline, without a simulation replay. Targeted phase demonstrations, 8.3% of the training frames, cover the suction transitions and the configurations an interrupted grasp leaves behind. On a real Unitree Z1, one multi-task checkpoint reaches 96.67% average success in both colour settings, 16.67 and 10.00 points above the strongest baseline, its monochromatic four-block successes averaging 15.92 mm of error. Four ablation settings fall 5.00 to 13.33 points short. We will release code and trained weights.",
    "history": "Submission history From: Yixian Zou [ view email ] [v1] Sat, 26 Sep 2026 16:40:23 UTC (4,109 KB)",
    "meta": {
      "citation_title": "CLAP: Closed-Loop Alignment with Pressure for Precise Suction Manipulation",
      "citation_author": "Liu, Shuaicheng",
      "citation_date": "2026/09/26",
      "citation_online_date": "2026/09/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.32767",
      "citation_arxiv_id": "2609.32767",
      "citation_abstract": "Stacking and palletising demand precise placement: error left in one layer is inherited by the next, and a flat pad offers no feature to funnel a wrong pose into the right one. Top-down suction suits such dense arrangements, and suction has already been brought into vision-language-action (VLA) policies. What that work does not report, however, is a policy conditioned on a measured vacuum signal, or one that uses it to abandon an action already under way. Vision does not settle the question here, because at the moment it matters the cup and the face it holds occlude each other. We present CLAP, which makes the attachment state observable through a pressure module tapped into the vacuum line. The decoded reading replaces the suction command in the policy's proprioception, is fused with the visual features, and terminates the open-loop execution window so that the policy re-infers from a fresh observation. For data, we record goal-state disassembly on the physical robot and reverse the joint-state sequence offline, without a simulation replay. Targeted phase demonstrations, 8.3% of the training frames, cover the suction transitions and the configurations an interrupted grasp leaves behind. On a real Unitree Z1, one multi-task checkpoint reaches 96.67% average success in both colour settings, 16.67 and 10.00 points above the strongest baseline, its monochromatic four-block successes averaging 15.92 mm of error. Four ablation settings fall 5.00 to 13.33 points short. We will release code and trained weights."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-27T00:40:23+08:00"
  },
  {
    "id": "2609.33269v1",
    "title": "Q-WAM: 4-Bit Quantization of World Action Models with Action-Subspace Protection",
    "abstract": "World Action Models (WAMs) jointly generate video and robot actions through iterative diffusion and perform strongly in robotic manipulation. However, their prohibitive compute and memory costs pose substantial deployment challenges. Post-training quantization (PTQ) can reduce these costs, but existing PTQ methods such as smoothing and rotation are insufficient to maintain the precision of action generation. To overcome this limitation, we propose Q-WAM, a new 4-bit weight-activation quantization for WAMs that preserves the actions the model generates. Specifically, we introduce the \\textit{Action Observability Gramian (AOG)}, which measures how much rounding errors in each weighted combination of a layer's input channels change the final action through all denoising steps. We also develop Action-Subspace Protection (ASP), which keeps the few most action-sensitive channel combinations in a tiny 16-bit low-rank branch and quantizes the complementary weights and activations to 4 bits, both as dense matrix multiplications that run efficiently on GPUs. Finally, to preserve action quality with minimal overhead, we identify the experts that matter most for the generated action by aggregating the AOG-derived action mass across the layers of each expert and apply ASP only to those experts. We evaluate Q-WAM on three WAMs, both in simulation and in real-world deployment. On the RoboTwin 2.0 benchmark, it reaches 89.6--93.0\\% average success rate, within 1.1 percentage points of the 16-bit models, while reducing the memory of the quantized blocks by 3.1--3.4$\\times$. Our method outperforms the strongest baseline, SVDQuant, by 2.5--8.7 percentage points. On a Unitree G1 humanoid and a bimanual UR3 robot, it improves success over SVDQuant by 12.8-17.6 percentage points.",
    "history": "Submission history From: Arash Akbari [ view email ] [v1] Sun, 27 Sep 2026 06:24:30 UTC (1,916 KB)",
    "meta": {
      "citation_title": "Q-WAM: 4-Bit Quantization of World Action Models with Action-Subspace Protection",
      "citation_author": "Wang, Yanzhi",
      "citation_date": "2026/09/27",
      "citation_online_date": "2026/09/27",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.33269",
      "citation_arxiv_id": "2609.33269",
      "citation_abstract": "World Action Models (WAMs) jointly generate video and robot actions through iterative diffusion and perform strongly in robotic manipulation. However, their prohibitive compute and memory costs pose substantial deployment challenges. Post-training quantization (PTQ) can reduce these costs, but existing PTQ methods such as smoothing and rotation are insufficient to maintain the precision of action generation. To overcome this limitation, we propose Q-WAM, a new 4-bit weight-activation quantization for WAMs that preserves the actions the model generates. Specifically, we introduce the \\textit{Action Observability Gramian (AOG)}, which measures how much rounding errors in each weighted combination of a layer's input channels change the final action through all denoising steps. We also develop Action-Subspace Protection (ASP), which keeps the few most action-sensitive channel combinations in a tiny 16-bit low-rank branch and quantizes the complementary weights and activations to 4 bits, both as dense matrix multiplications that run efficiently on GPUs. Finally, to preserve action quality with minimal overhead, we identify the experts that matter most for the generated action by aggregating the AOG-derived action mass across the layers of each expert and apply ASP only to those experts. We evaluate Q-WAM on three WAMs, both in simulation and in real-world deployment. On the RoboTwin 2.0 benchmark, it reaches 89.6--93.0\\% average success rate, within 1.1 percentage points of the 16-bit models, while reducing the memory of the quantized blocks by 3.1--3.4$\\times$. Our method outperforms the strongest baseline, SVDQuant, by 2.5--8.7 percentage points. On a Unitree G1 humanoid and a bimanual UR3 robot, it improves success over SVDQuant by 12.8-17.6 percentage points."
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-27T14:24:30+08:00"
  },
  {
    "id": "2609.33310v1",
    "title": "CompliantWBC: Whole-Body Compliance for Heavy Humanoids via Force Latent Estimation and Residual Impedance Targets",
    "abstract": "Whole-body compliant control is essential for deploying heavy humanoids under high payload in human-centric environments. Most prior force-aware learning-based pipelines focus on end-effector resistance, per-link upper-body springs, or end-effector stiffness modulation, leaving arbitrary-site perturbations on heavy platforms with lower-body engagement largely unaddressed. We close this gap with CompliantWBC comprising: (1) A base policy trained with RL to maximize compliance-fidelity reward, guided by a multi-site whole-body impedance reference controller, extending classical Cartesian impedance to any controlled link; (2) A bounded residual policy that edits the per-link impedance equilibrium over a frozen base, correcting the coarse but structured wrench estimate supplied by a force encoder co-trained behind a gradient barrier; (3) A Phong-weighted force-origin sampler with an axis-decoupled pelvis anchor induces lower-body-inclusive compliance curriculum training via two interpretable parameters. We evaluate CompliantWBC in simulation against both compliant and stiff baselines, achieving best compliant fidelity of 2.58cm deviation from analytical solutions, and demonstrate it on a real heavy humanoid across static/dynamic force reaction, board wiping, squat under payload, and cooperative payload transport. Project website: this https URL",
    "history": "Submission history From: Tuan Dat Phuong [ view email ] [v1] Sun, 27 Sep 2026 07:24:14 UTC (1,875 KB)",
    "meta": {
      "citation_title": "CompliantWBC: Whole-Body Compliance for Heavy Humanoids via Force Latent Estimation and Residual Impedance Targets",
      "citation_author": "Le, An Thai",
      "citation_date": "2026/09/27",
      "citation_online_date": "2026/09/27",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.33310",
      "citation_arxiv_id": "2609.33310",
      "citation_abstract": "Whole-body compliant control is essential for deploying heavy humanoids under high payload in human-centric environments. Most prior force-aware learning-based pipelines focus on end-effector resistance, per-link upper-body springs, or end-effector stiffness modulation, leaving arbitrary-site perturbations on heavy platforms with lower-body engagement largely unaddressed. We close this gap with CompliantWBC comprising: (1) A base policy trained with RL to maximize compliance-fidelity reward, guided by a multi-site whole-body impedance reference controller, extending classical Cartesian impedance to any controlled link; (2) A bounded residual policy that edits the per-link impedance equilibrium over a frozen base, correcting the coarse but structured wrench estimate supplied by a force encoder co-trained behind a gradient barrier; (3) A Phong-weighted force-origin sampler with an axis-decoupled pelvis anchor induces lower-body-inclusive compliance curriculum training via two interpretable parameters. We evaluate CompliantWBC in simulation against both compliant and stiff baselines, achieving best compliant fidelity of 2.58cm deviation from analytical solutions, and demonstrate it on a real heavy humanoid across static/dynamic force reaction, board wiping, squat under payload, and cooperative payload transport. Project website: https://dotandung.github.io/compliantwbc/"
    },
    "groups": [
      "w3"
    ],
    "event_at": "2026-09-27T15:24:14+08:00"
  },
  {
    "id": "2609.37771v1",
    "title": "Faster and Better? Benchmark Bugs and Design Limitations Distort the Evaluation of Vision-Language-Action Acceleration",
    "abstract": "Simulated manipulation benchmarks are the standard tool for evaluating vision-language-action (VLA) policies and the acceleration methods that reduce their inference latency for on-robot deployment. On these benchmarks, we observe that some training-free acceleration methods, which approximate the baseline policy's computation, achieve higher measured success rates than the baseline itself. Success rates alone cannot establish whether such gains come from better task execution or from evaluation flaws. We therefore investigate two kinds of benchmark flaws behind these gains: bugs, where the implementation does not match the intended task or evaluation protocol, and design limitations, where success criteria and simulation settings do not fully capture how acceleration affects task execution. Starting from tasks with anomalous gains, we localize root causes by plotting object trajectories against checker acceptance regions, and classify the resulting bugs into task consistency, initialization, and reproducibility. Extending this audit to seven benchmarks, including RoboTwin, LIBERO-Plus, and VLABench, we identify 22 bugs of these types and 4 design limitations. For the latter, we revise permissive success checkers, correct unrealistic object masses, and add a motion-aware score that favors smoother actions. Experiments show that bug fixes can reverse method rankings, moving the baseline from last to first on one task. Addressing design limitations can likewise remove anomalous gains: on another task, the baseline moves from 21 percentage points behind an accelerated method to 5 points ahead. Gains attributed to acceleration can therefore be artifacts of the benchmark rather than better task execution. We release our bug fixes and revised benchmark settings to support trustworthy evaluation of VLA acceleration.",
    "history": "Submission history From: Qiwei Chen [ view email ] [v1] Tue, 29 Sep 2026 15:12:55 UTC (5,266 KB)",
    "meta": {
      "citation_title": "Faster and Better? Benchmark Bugs and Design Limitations Distort the Evaluation of Vision-Language-Action Acceleration",
      "citation_author": "Gu, Jinyu",
      "citation_date": "2026/09/29",
      "citation_online_date": "2026/09/29",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.37771",
      "citation_arxiv_id": "2609.37771",
      "citation_abstract": "Simulated manipulation benchmarks are the standard tool for evaluating vision-language-action (VLA) policies and the acceleration methods that reduce their inference latency for on-robot deployment. On these benchmarks, we observe that some training-free acceleration methods, which approximate the baseline policy's computation, achieve higher measured success rates than the baseline itself. Success rates alone cannot establish whether such gains come from better task execution or from evaluation flaws. We therefore investigate two kinds of benchmark flaws behind these gains: bugs, where the implementation does not match the intended task or evaluation protocol, and design limitations, where success criteria and simulation settings do not fully capture how acceleration affects task execution. Starting from tasks with anomalous gains, we localize root causes by plotting object trajectories against checker acceptance regions, and classify the resulting bugs into task consistency, initialization, and reproducibility. Extending this audit to seven benchmarks, including RoboTwin, LIBERO-Plus, and VLABench, we identify 22 bugs of these types and 4 design limitations. For the latter, we revise permissive success checkers, correct unrealistic object masses, and add a motion-aware score that favors smoother actions. Experiments show that bug fixes can reverse method rankings, moving the baseline from last to first on one task. Addressing design limitations can likewise remove anomalous gains: on another task, the baseline moves from 21 percentage points behind an accelerated method to 5 points ahead. Gains attributed to acceleration can therefore be artifacts of the benchmark rather than better task execution. We release our bug fixes and revised benchmark settings to support trustworthy evaluation of VLA acceleration."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-09-29T23:12:55+08:00"
  },
  {
    "id": "2609.38172v1",
    "title": "Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation",
    "abstract": "Teaching humanoids loco-manipulation skills, such as carrying diverse objects, via visual imitation is a promising path toward generalist robots. However, collecting diverse, high-quality interaction videos, such as clips that clearly show a person's full body and unoccluded interactions with objects, poses a practical barrier to scaling this approach. We propose PRISM, a real-to-sim-to-real framework that overcomes this limitation by amplifying a handful of real videos into a large, diverse training set. PRISM first generates hundreds of diverse \"counterfactual\" human-object interaction videos via video-to-video (V2V) generation from a few exemplar real videos. Our contact-anchored real-to-sim pipeline then reconstructs both human and object motions, retargeting this imperfect video data into physically plausible trajectories. The intra-class variability across these counterfactual videos lets us train a single policy that generalizes to unseen objects within each category. We demonstrate the full pipeline by deploying this policy on a real robot without any real-world fine-tuning. Using only onboard depth observations, our humanoid picks up, carries, and drops objects, including boxes, barrels, bins, and balls, across novel instances, sizes, and initial configurations.",
    "history": "Submission history From: Zihan Wang [ view email ] [v1] Tue, 29 Sep 2026 17:59:45 UTC (31,240 KB)",
    "meta": {
      "citation_title": "Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation",
      "citation_author": "Kanazawa, Angjoo",
      "citation_date": "2026/09/29",
      "citation_online_date": "2026/09/29",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.38172",
      "citation_arxiv_id": "2609.38172",
      "citation_abstract": "Teaching humanoids loco-manipulation skills, such as carrying diverse objects, via visual imitation is a promising path toward generalist robots. However, collecting diverse, high-quality interaction videos, such as clips that clearly show a person's full body and unoccluded interactions with objects, poses a practical barrier to scaling this approach. We propose PRISM, a real-to-sim-to-real framework that overcomes this limitation by amplifying a handful of real videos into a large, diverse training set. PRISM first generates hundreds of diverse \"counterfactual\" human-object interaction videos via video-to-video (V2V) generation from a few exemplar real videos. Our contact-anchored real-to-sim pipeline then reconstructs both human and object motions, retargeting this imperfect video data into physically plausible trajectories. The intra-class variability across these counterfactual videos lets us train a single policy that generalizes to unseen objects within each category. We demonstrate the full pipeline by deploying this policy on a real robot without any real-world fine-tuning. Using only onboard depth observations, our humanoid picks up, carries, and drops objects, including boxes, barrels, bins, and balls, across novel instances, sizes, and initial configurations."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-09-30T01:59:45+08:00"
  },
  {
    "id": "2609.39179v1",
    "title": "LocoWM: High-Precision Locomotion through World-Model-Guided Residual Adaptation",
    "abstract": "High-precision locomotion combines motion-command tracking with precise regulation of task-relevant physical states, enabling robots to interact reliably with their surroundings during motion. Joint end-to-end optimization can leave precision objectives insufficiently optimized, while reactive residual control adjusts actions only after deviations become observable. We present \\textbf{LocoWM}, a world-model-guided preactive residual adaptation framework for high-precision locomotion. A base policy provides command-following locomotion, while an action-conditioned world model predicts a sequence of future physical states from proprioceptive history and the proposed base action. A residual adapter conditions on this predicted sequence to generate additive action corrections that compensate for anticipated deviations. Two-stage training first learns locomotion and action-conditioned dynamics, then freezes both modules while training the adapter, separating locomotion acquisition from precision adaptation. Experiments spanning terrain leveling, acceleration compensation, and push recovery demonstrate improved control precision and disturbance robustness over end-to-end and reactive residual baselines. Demos and code are available at: this https URL",
    "history": "Submission history From: Zijie Zhao [ view email ] [v1] Wed, 30 Sep 2026 07:35:08 UTC (11,326 KB)",
    "meta": {
      "citation_title": "LocoWM: High-Precision Locomotion through World-Model-Guided Residual Adaptation",
      "citation_author": "Zhao, Dongbin",
      "citation_date": "2026/09/30",
      "citation_online_date": "2026/09/30",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.39179",
      "citation_arxiv_id": "2609.39179",
      "citation_abstract": "High-precision locomotion combines motion-command tracking with precise regulation of task-relevant physical states, enabling robots to interact reliably with their surroundings during motion. Joint end-to-end optimization can leave precision objectives insufficiently optimized, while reactive residual control adjusts actions only after deviations become observable. We present \\textbf{LocoWM}, a world-model-guided preactive residual adaptation framework for high-precision locomotion. A base policy provides command-following locomotion, while an action-conditioned world model predicts a sequence of future physical states from proprioceptive history and the proposed base action. A residual adapter conditions on this predicted sequence to generate additive action corrections that compensate for anticipated deviations. Two-stage training first learns locomotion and action-conditioned dynamics, then freezes both modules while training the adapter, separating locomotion acquisition from precision adaptation. Experiments spanning terrain leveling, acceleration compensation, and push recovery demonstrate improved control precision and disturbance robustness over end-to-end and reactive residual baselines. Demos and code are available at: https://zhaozijie2022.github.io/LocoWM"
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-09-30T15:35:08+08:00"
  },
  {
    "id": "2610.03997v1",
    "title": "REDIRECT: A 1% Fix for Bad Robot Habits",
    "abstract": "Robots can acquire bad habits from a few defective moments in otherwise useful teleoperation. In mixed-quality robot data, normal and problematic demonstrations share most task behavior and differ only at a local action continuation. Full retraining is costly, while fine-tuning on clean data alone offers limited recovery under a small update budget. We ask whether the local difference can instead be fixed with only 1% of the full-training sample-backward budget. Using only episode-level retained/problematic labels, REDIRECT localizes the branch, assigns a coherent retained continuation to the original observations around it, and anchors shared behavior, without frame-level annotations or additional interaction. Across three randomized ManiSkill tasks and three seeds, REDIRECT raises the mean success rate from 68.2% to 90.3%, recovering 88.8% of the clean-retraining gap versus 22.1% for matched-compute fine-tuning. On a PiPER arm, it recovers 86.7-92.9% of the clean-only gap across cup insertion and towel folding. Local robot habits can therefore be repaired by spending the update budget on the behavioral difference rather than relearning shared behavior.",
    "history": "Submission history From: Jiazhuo Li [ view email ] [v1] Fri, 2 Oct 2026 20:00:22 UTC (2,354 KB)",
    "meta": {
      "citation_title": "REDIRECT: A 1% Fix for Bad Robot Habits",
      "citation_author": "Liu, Houde",
      "citation_date": "2026/10/02",
      "citation_online_date": "2026/10/02",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.03997",
      "citation_arxiv_id": "2610.03997",
      "citation_abstract": "Robots can acquire bad habits from a few defective moments in otherwise useful teleoperation. In mixed-quality robot data, normal and problematic demonstrations share most task behavior and differ only at a local action continuation. Full retraining is costly, while fine-tuning on clean data alone offers limited recovery under a small update budget. We ask whether the local difference can instead be fixed with only 1% of the full-training sample-backward budget. Using only episode-level retained/problematic labels, REDIRECT localizes the branch, assigns a coherent retained continuation to the original observations around it, and anchors shared behavior, without frame-level annotations or additional interaction. Across three randomized ManiSkill tasks and three seeds, REDIRECT raises the mean success rate from 68.2% to 90.3%, recovering 88.8% of the clean-retraining gap versus 22.1% for matched-compute fine-tuning. On a PiPER arm, it recovers 86.7-92.9% of the clean-only gap across cup insertion and towel folding. Local robot habits can therefore be repaired by spending the update budget on the behavioral difference rather than relearning shared behavior."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-10-03T04:00:22+08:00"
  },
  {
    "id": "2610.04009v1",
    "title": "SUAVE: Unified Video-Action Models via Masked Diffusion",
    "abstract": "Vision-language-action models (VLAs) inherit strong semantic grounding from pretrained vision-language backbones but are typically optimized for predicting actions rather than future observations. They can see and act, but they do not imagine the future before acting. World action models (WAMs) built on video diffusion backbones can imagine but treat language as frozen conditioning on a continuous latent space. Unified models bring these modalities into one architecture, but they either decode autoregressively, one token at a time, or keep video continuous with an auxiliary action head. In this work, we present SUAVE, a Single vocabulary Unified Action-Video modEl in which a masked diffusion transformer generates video and actions conditioned on language, with all three modalities represented as discrete tokens in a shared sequence. Choosing which tokens to mask at inference turns the same network into a world model, a robot policy, or a video-action model. For action-free co-training, the action positions of unlabeled video are filled with mask tokens and excluded from the loss. Simulation and real-world experiments demonstrate two findings. First, a single SUAVE model predicts long-horizon video and acts as a policy, competitive with dedicated world models and specialized action policies on static and dynamic manipulation tasks. On a real robot, our model generates subgoal images and an action chunk spanning one second of motion in 1,030 ms on an RTX 5090 GPU, sustaining closed-loop control at 2.5 actions per second. Second, pretraining on robot video and co-training on human video substantially improves policy performance and zero-shot robustness to distribution shift. Together, these results show that masked diffusion is a practical and versatile foundation for unified video-action modeling.",
    "history": "Submission history From: Rhythm Syed [ view email ] [v1] Fri, 2 Oct 2026 20:11:13 UTC (7,941 KB)",
    "meta": {
      "citation_title": "SUAVE: Unified Video-Action Models via Masked Diffusion",
      "citation_author": "Dear, Tony",
      "citation_date": "2026/10/02",
      "citation_online_date": "2026/10/02",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.04009",
      "citation_arxiv_id": "2610.04009",
      "citation_abstract": "Vision-language-action models (VLAs) inherit strong semantic grounding from pretrained vision-language backbones but are typically optimized for predicting actions rather than future observations. They can see and act, but they do not imagine the future before acting. World action models (WAMs) built on video diffusion backbones can imagine but treat language as frozen conditioning on a continuous latent space. Unified models bring these modalities into one architecture, but they either decode autoregressively, one token at a time, or keep video continuous with an auxiliary action head. In this work, we present SUAVE, a Single vocabulary Unified Action-Video modEl in which a masked diffusion transformer generates video and actions conditioned on language, with all three modalities represented as discrete tokens in a shared sequence. Choosing which tokens to mask at inference turns the same network into a world model, a robot policy, or a video-action model. For action-free co-training, the action positions of unlabeled video are filled with mask tokens and excluded from the loss. Simulation and real-world experiments demonstrate two findings. First, a single SUAVE model predicts long-horizon video and acts as a policy, competitive with dedicated world models and specialized action policies on static and dynamic manipulation tasks. On a real robot, our model generates subgoal images and an action chunk spanning one second of motion in 1,030 ms on an RTX 5090 GPU, sustaining closed-loop control at 2.5 actions per second. Second, pretraining on robot video and co-training on human video substantially improves policy performance and zero-shot robustness to distribution shift. Together, these results show that masked diffusion is a practical and versatile foundation for unified video-action modeling."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-10-03T04:11:13+08:00"
  },
  {
    "id": "2610.04231v1",
    "title": "Continual Humanoid Motion Learning",
    "abstract": "Humanoid whole-body controllers can now track a diverse set of dynamic motions, but they are typically trained offline and then frozen, so teaching such a controller a new skill tends to erode the skills it already mastered. We study continual learning for humanoid whole-body motion, where a single controller must acquire skills from a sequential task stream without revisiting past data. We introduce Similarity-guided LoRA-PNN, a progressive neural network (PNN) policy that prevents catastrophic forgetting by construction while reusing knowledge across skills through lightweight low-rank adaptation. A two-level motion-similarity measure, built from dynamic time warping aggregated by optimal transport, decides which prior skill to build on and how much new capacity to allocate, yielding strong forward transfer and large efficiency gains. Across six sequentially learned skill categories, our similarity-guided LoRA policy attains the best forward transfer (0.125 vs. 0.079) and the highest average accuracy among all methods, while saving up to 94.5% of trainable parameters and 40.8% of training time. The resulting controller reaches 96.13% sim-to-sim transfer and is deployed on a physical Unitree G1. Our code is available at this https URL .",
    "history": "Submission history From: Zhewen He [ view email ] [v1] Sat, 3 Oct 2026 02:45:37 UTC (16,972 KB)",
    "meta": {
      "citation_title": "Continual Humanoid Motion Learning",
      "citation_author": "Fang, Yi",
      "citation_date": "2026/10/03",
      "citation_online_date": "2026/10/03",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.04231",
      "citation_arxiv_id": "2610.04231",
      "citation_abstract": "Humanoid whole-body controllers can now track a diverse set of dynamic motions, but they are typically trained offline and then frozen, so teaching such a controller a new skill tends to erode the skills it already mastered. We study continual learning for humanoid whole-body motion, where a single controller must acquire skills from a sequential task stream without revisiting past data. We introduce Similarity-guided LoRA-PNN, a progressive neural network (PNN) policy that prevents catastrophic forgetting by construction while reusing knowledge across skills through lightweight low-rank adaptation. A two-level motion-similarity measure, built from dynamic time warping aggregated by optimal transport, decides which prior skill to build on and how much new capacity to allocate, yielding strong forward transfer and large efficiency gains. Across six sequentially learned skill categories, our similarity-guided LoRA policy attains the best forward transfer (0.125 vs. 0.079) and the highest average accuracy among all methods, while saving up to 94.5% of trainable parameters and 40.8% of training time. The resulting controller reaches 96.13% sim-to-sim transfer and is deployed on a physical Unitree G1. Our code is available at https://anonymous.4open.science/r/continual-humanoid-learning-35D3."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-10-03T10:45:37+08:00"
  },
  {
    "id": "2610.04363v1",
    "title": "TacOT: Learning Contact-Rich Dexterous Manipulation from Human Demonstrations via Tactile-Guided Optimal Transport",
    "abstract": "Learning contact-rich dexterous manipulation from human demonstrations provides a scalable source of interaction data, yet transferring such skills to robots remains challenging due to unreliable human--robot correspondence. Existing human-to-robot transfer methods typically rely on visual appearance or motion similarity, which may associate similar motions with different contact states and force patterns. Tactile dynamics provide interaction-aware cues to distinguish manipulation processes with similar motions but different contact states. We introduce tactile-guided optimal transport (TacOT), a framework for human-to-robot contact-rich manipulation. TacOT leverages action--tactile dynamic time warping to identify human--robot demonstration correspondences with consistent interaction dynamics and uses these correspondences to guide soft optimal transport alignment in a shared policy representation space. This enables human demonstrations to provide contact-rich supervision for robot policy learning without requiring predefined frame-level human--robot pairing. Across four real-world dexterous manipulation tasks, TacOT improves closed-loop success rates over action-guided OT by up to 17 points on in-distribution tasks and 20 points under targeted human-to-robot out-of-distribution transfer. Further analyses show that tactile-guided correspondence selects demonstration pairs with more consistent contact dynamics and produces latent representations that better reflect interaction-state evolution. These results demonstrate that tactile dynamics provide an effective semantic signal for establishing reliable human-to-robot correspondence in contact-rich dexterous manipulation.",
    "history": "Submission history From: Xingting Li [ view email ] [v1] Sat, 3 Oct 2026 08:00:56 UTC (24,595 KB)",
    "meta": {
      "citation_title": "TacOT: Learning Contact-Rich Dexterous Manipulation from Human Demonstrations via Tactile-Guided Optimal Transport",
      "citation_author": "Ding, Wenbo",
      "citation_date": "2026/10/03",
      "citation_online_date": "2026/10/03",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.04363",
      "citation_arxiv_id": "2610.04363",
      "citation_abstract": "Learning contact-rich dexterous manipulation from human demonstrations provides a scalable source of interaction data, yet transferring such skills to robots remains challenging due to unreliable human--robot correspondence. Existing human-to-robot transfer methods typically rely on visual appearance or motion similarity, which may associate similar motions with different contact states and force patterns. Tactile dynamics provide interaction-aware cues to distinguish manipulation processes with similar motions but different contact states. We introduce tactile-guided optimal transport (TacOT), a framework for human-to-robot contact-rich manipulation. TacOT leverages action--tactile dynamic time warping to identify human--robot demonstration correspondences with consistent interaction dynamics and uses these correspondences to guide soft optimal transport alignment in a shared policy representation space. This enables human demonstrations to provide contact-rich supervision for robot policy learning without requiring predefined frame-level human--robot pairing. Across four real-world dexterous manipulation tasks, TacOT improves closed-loop success rates over action-guided OT by up to 17 points on in-distribution tasks and 20 points under targeted human-to-robot out-of-distribution transfer. Further analyses show that tactile-guided correspondence selects demonstration pairs with more consistent contact dynamics and produces latent representations that better reflect interaction-state evolution. These results demonstrate that tactile dynamics provide an effective semantic signal for establishing reliable human-to-robot correspondence in contact-rich dexterous manipulation."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-10-03T16:00:56+08:00"
  },
  {
    "id": "2610.05062v1",
    "title": "Beyond LLM Serving: Characterizing Vision-Language-Action Workloads for Embodied AI System Design",
    "abstract": "Vision-language-action (VLA) models translate multimodal observations into low-level robot actions. During robot operation, each control period sets an inference deadline, and overruns leave the robot acting on stale observations, reducing task success. Meeting this deadline motivates on-device or nearby edge execution, where a single robot requires batch-1 inference outside the design point of LLM serving systems. Although VLA architectures combine familiar vision-language, autoregressive, and diffusion-style components, their runtime behavior in this batch-1 control setting remains uncharacterized. We characterize four representative VLA models on an edge GPU server and two onboard SoCs, using single-inference profiling and 43,200 closed-loop episodes. Action tensor dimensionality determines whether a stage is memory- or compute-bound, platform balance can shift that bottleneck, and GPU frequency scaling yields a platform-dependent energy-latency sweet spot. In closed-loop operation, overlapping inference with action execution creates an accuracy-speed-energy tradeoff, and no configuration is Pareto-dominant across deployment SLOs. These results guide joint design of VLA model architectures, hardware, and runtime policies.",
    "history": "Submission history From: Seonghun Jung [ view email ] [v1] Sun, 4 Oct 2026 09:05:46 UTC (714 KB)",
    "meta": {
      "citation_title": "Beyond LLM Serving: Characterizing Vision-Language-Action Workloads for Embodied AI System Design",
      "citation_author": "Huh, Jaehyuk",
      "citation_doi": "10.1145/3845814.3850089",
      "citation_date": "2026/10/04",
      "citation_online_date": "2026/10/04",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.05062",
      "citation_arxiv_id": "2610.05062",
      "citation_abstract": "Vision-language-action (VLA) models translate multimodal observations into low-level robot actions. During robot operation, each control period sets an inference deadline, and overruns leave the robot acting on stale observations, reducing task success. Meeting this deadline motivates on-device or nearby edge execution, where a single robot requires batch-1 inference outside the design point of LLM serving systems. Although VLA architectures combine familiar vision-language, autoregressive, and diffusion-style components, their runtime behavior in this batch-1 control setting remains uncharacterized. We characterize four representative VLA models on an edge GPU server and two onboard SoCs, using single-inference profiling and 43,200 closed-loop episodes. Action tensor dimensionality determines whether a stage is memory- or compute-bound, platform balance can shift that bottleneck, and GPU frequency scaling yields a platform-dependent energy-latency sweet spot. In closed-loop operation, overlapping inference with action execution creates an accuracy-speed-energy tradeoff, and no configuration is Pareto-dominant across deployment SLOs. These results guide joint design of VLA model architectures, hardware, and runtime policies."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-10-04T17:05:46+08:00"
  },
  {
    "id": "2610.05414v1",
    "title": "RMMBench: A Comprehensive Benchmark for Robotic Mobile Manipulation",
    "abstract": "Although the advancement of vision-language models (VLMs) has endowed robots with enhanced environmental understanding and task reasoning, a comprehensive evaluation methodology is important to advance the integration of VLMs in robotic navigation and manipulation. However, current benchmarks lack a comprehensive method to evaluate diverse robotic tasks, and evaluation metrics remain relatively constrained, making it difficult to assess the embodied capabilities of VLMs in a thorough and fine-grained manner. To address this issue, we propose RMMBench, an evaluation benchmark that requires robots to understand language instructions and perform long-horizon tasks in continuous spaces. RMMBench seamlessly integrates high- and low-level embodied tasks into a unified framework, constructing a \"navigation-manipulation\" task suite comprising 70 canonical task scenarios that range from localized manipulation to long-horizon composite navigation. The results reveal that leading VLMs still face major challenges in spatial localization when performing mobile manipulation tasks, and also highlight the necessity of enhancing the spatial perception capability of robots during long-horizon interactions. RMMBench can be accessed at this https URL",
    "history": "Submission history From: Huapeng Li [ view email ] [v1] Sun, 4 Oct 2026 17:55:26 UTC (24,181 KB)",
    "meta": {
      "citation_title": "RMMBench: A Comprehensive Benchmark for Robotic Mobile Manipulation",
      "citation_author": "Zhang, Wei",
      "citation_date": "2026/10/04",
      "citation_online_date": "2026/10/04",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.05414",
      "citation_arxiv_id": "2610.05414",
      "citation_abstract": "Although the advancement of vision-language models (VLMs) has endowed robots with enhanced environmental understanding and task reasoning, a comprehensive evaluation methodology is important to advance the integration of VLMs in robotic navigation and manipulation. However, current benchmarks lack a comprehensive method to evaluate diverse robotic tasks, and evaluation metrics remain relatively constrained, making it difficult to assess the embodied capabilities of VLMs in a thorough and fine-grained manner. To address this issue, we propose RMMBench, an evaluation benchmark that requires robots to understand language instructions and perform long-horizon tasks in continuous spaces. RMMBench seamlessly integrates high- and low-level embodied tasks into a unified framework, constructing a \"navigation-manipulation\" task suite comprising 70 canonical task scenarios that range from localized manipulation to long-horizon composite navigation. The results reveal that leading VLMs still face major challenges in spatial localization when performing mobile manipulation tasks, and also highlight the necessity of enhancing the spatial perception capability of robots during long-horizon interactions. RMMBench can be accessed at https://mxxq-stack.github.io/rmmbench-project/"
    },
    "groups": [],
    "event_at": "2026-10-05T01:55:26+08:00",
    "exclusion": "UTC date falls in the following Beijing week; not selected for the closed week"
  },
  {
    "id": "2610.05407v1",
    "title": "TUCO: Curating Simulation Demonstrations for Sim-to-Real Robot Policy Co-Training",
    "abstract": "Simulation demonstrations can supplement scarce real-world data for robot policy co-training. However, the value of using data curation to actively select these demonstrations for sim-to-real co-training remains underexplored. Existing curation methods also lack a unified criterion for measuring trajectory-level utility and set-level coverage from closed-loop target behavior. To address these gaps, we present the first systematic study of data curation for sim-to-real robot policy co-training and propose Trajectory-level Utility and set-level Coverage Optimization (TUCO). TUCO uses influence functions to trace how each source demonstration affects target-domain scoring rollouts. Our key insight is that these effects can be decomposed into an overall contribution to target return and variation across rollouts, providing a common closed-loop basis for measuring trajectory utility and set coverage. We further propose a performance-aligned subset optimizer that combines these measures in a unified curation objective to reduce redundancy and select complementary demonstrations. Extensive experiments on RoboMimic and OmniReset establish the value of active simulation data curation for sim-to-real policy co-training and show that TUCO achieves state-of-the-art performance across single-simulator, sim-to-sim, and sim-to-real settings.",
    "history": "Submission history From: Mengfei Zhao [ view email ] [v1] Sun, 4 Oct 2026 17:45:50 UTC (7,699 KB)",
    "meta": {
      "citation_title": "TUCO: Curating Simulation Demonstrations for Sim-to-Real Robot Policy Co-Training",
      "citation_author": "Yang, Jianfei",
      "citation_date": "2026/10/04",
      "citation_online_date": "2026/10/04",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.05407",
      "citation_arxiv_id": "2610.05407",
      "citation_abstract": "Simulation demonstrations can supplement scarce real-world data for robot policy co-training. However, the value of using data curation to actively select these demonstrations for sim-to-real co-training remains underexplored. Existing curation methods also lack a unified criterion for measuring trajectory-level utility and set-level coverage from closed-loop target behavior. To address these gaps, we present the first systematic study of data curation for sim-to-real robot policy co-training and propose Trajectory-level Utility and set-level Coverage Optimization (TUCO). TUCO uses influence functions to trace how each source demonstration affects target-domain scoring rollouts. Our key insight is that these effects can be decomposed into an overall contribution to target return and variation across rollouts, providing a common closed-loop basis for measuring trajectory utility and set coverage. We further propose a performance-aligned subset optimizer that combines these measures in a unified curation objective to reduce redundancy and select complementary demonstrations. Extensive experiments on RoboMimic and OmniReset establish the value of active simulation data curation for sim-to-real policy co-training and show that TUCO achieves state-of-the-art performance across single-simulator, sim-to-sim, and sim-to-real settings."
    },
    "groups": [],
    "event_at": "2026-10-05T01:45:50+08:00",
    "exclusion": "UTC date falls in the following Beijing week; not selected for the closed week"
  },
  {
    "id": "2610.09943v1",
    "title": "Many Ways to Succeed: Diversity-Driven RL Fine-Tuning for VLA Generalization",
    "abstract": "Reinforcement learning (RL) fine-tuning improves vision-language-action (VLA) policies through closed-loop experience, yet generalization beyond the fine-tuning distribution remains limited. Our analysis reveals a selective reshaping of exploration: RL contracts behavior globally, yet diversifies successful trajectories, elicits success with fewer rollouts, and covers more of the latent task-valid solution space than supervised fine-tuning. Broader successful-mode coverage may provide alternative strategies under distribution shifts. Inspired by this, we introduce DRIVE (Diversity-driven RL fIne-tuning for VLA gEneralization), which turns successful-behavior diversity into an explicit RL objective. DRIVE groups rollouts under matched task conditions, compares their trajectories with temporal alignment, and derives a success-conditioned intrinsic reward from relative behavioral diversity. This design encourages broader coverage of feasible solutions without rewarding diverse failures or superficial timing differences. Across LIBERO-Plus, ManiSkill3, and RoboTwin 2.0, DRIVE improves the average out-of-domain (OOD) performance over vanilla RL fine-tuning by 5.3 points on $\\pi_0$ and 2.0 points on $\\pi_{0.5}$. On a dual-arm AgileX PiPER-X platform, DRIVE further increases average OOD success from 64.1% to 73.3% (+9.2 points), demonstrating gains that persist under physical deployment.",
    "history": "Submission history From: Haoru Li [ view email ] [v1] Wed, 7 Oct 2026 12:26:21 UTC (9,826 KB)",
    "meta": {
      "citation_title": "Many Ways to Succeed: Diversity-Driven RL Fine-Tuning for VLA Generalization",
      "citation_author": "Wang, Zhi",
      "citation_date": "2026/10/07",
      "citation_online_date": "2026/10/07",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.09943",
      "citation_arxiv_id": "2610.09943",
      "citation_abstract": "Reinforcement learning (RL) fine-tuning improves vision-language-action (VLA) policies through closed-loop experience, yet generalization beyond the fine-tuning distribution remains limited. Our analysis reveals a selective reshaping of exploration: RL contracts behavior globally, yet diversifies successful trajectories, elicits success with fewer rollouts, and covers more of the latent task-valid solution space than supervised fine-tuning. Broader successful-mode coverage may provide alternative strategies under distribution shifts. Inspired by this, we introduce DRIVE (Diversity-driven RL fIne-tuning for VLA gEneralization), which turns successful-behavior diversity into an explicit RL objective. DRIVE groups rollouts under matched task conditions, compares their trajectories with temporal alignment, and derives a success-conditioned intrinsic reward from relative behavioral diversity. This design encourages broader coverage of feasible solutions without rewarding diverse failures or superficial timing differences. Across LIBERO-Plus, ManiSkill3, and RoboTwin 2.0, DRIVE improves the average out-of-domain (OOD) performance over vanilla RL fine-tuning by 5.3 points on $\\pi_0$ and 2.0 points on $\\pi_{0.5}$. On a dual-arm AgileX PiPER-X platform, DRIVE further increases average OOD success from 64.1% to 73.3% (+9.2 points), demonstrating gains that persist under physical deployment."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-07T20:26:21+08:00"
  },
  {
    "id": "2610.12468v1",
    "title": "DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training",
    "abstract": "We present DreamTrue, a multi-view, cross-embodiment robot world model for action-faithful and physically plausible video prediction. Training such a model on existing robot datasets faces two obstacles: imprecise calibration can impair action following, while limited coverage of unsuccessful interactions can bias predictions toward successful outcomes. To improve action following across embodiments, we render action trajectories into image-space conditions and introduce offline geometric calibration to align these conditions with the target videos. To broaden interaction coverage, we introduce counterfactual post-training, modifying recorded action trajectories and generating future videos under a wider range of actions and contact configurations. To provide feedback on these predictions without paired ground-truth futures, we construct a human-annotated video dataset covering robot, object, and interaction defects and use it to train an embodied video reward model. Its scores guide reinforcement-learning post-training toward more physically plausible interaction outcomes. On AgiBot, DreamTrue attains state-of-the-art action following, while reducing the human-assessed interaction defect rate from from 48.12% to 6.25%. Notably, our model ranks first in the world model track of the AgiBot World Challenge 2026. The project page can be found at this https URL .",
    "history": "Submission history From: Lue Fan [ view email ] [v1] Thu, 8 Oct 2026 17:59:51 UTC (25,713 KB)",
    "meta": {
      "citation_title": "DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training",
      "citation_author": "Zhang, Zhaoxiang",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12468",
      "citation_arxiv_id": "2610.12468",
      "citation_abstract": "We present DreamTrue, a multi-view, cross-embodiment robot world model for action-faithful and physically plausible video prediction. Training such a model on existing robot datasets faces two obstacles: imprecise calibration can impair action following, while limited coverage of unsuccessful interactions can bias predictions toward successful outcomes. To improve action following across embodiments, we render action trajectories into image-space conditions and introduce offline geometric calibration to align these conditions with the target videos. To broaden interaction coverage, we introduce counterfactual post-training, modifying recorded action trajectories and generating future videos under a wider range of actions and contact configurations. To provide feedback on these predictions without paired ground-truth futures, we construct a human-annotated video dataset covering robot, object, and interaction defects and use it to train an embodied video reward model. Its scores guide reinforcement-learning post-training toward more physically plausible interaction outcomes. On AgiBot, DreamTrue attains state-of-the-art action following, while reducing the human-assessed interaction defect rate from from 48.12% to 6.25%. Notably, our model ranks first in the world model track of the AgiBot World Challenge 2026. The project page can be found at https://brave-eai.github.io/DreamTrue."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-09T01:59:51+08:00"
  },
  {
    "id": "2610.12470v1",
    "title": "Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration",
    "abstract": "While learning dexterous manipulation from a single human video offers a promising alternative to costly robot demonstrations, many recent methods predominantly imitate demonstrated motions. Such strict motion matching often limits generalization to initial object poses, goal poses, and grasps not shown in the video. Alternatively, discovering a policy via reinforcement learning (RL) allows for broad generalization, but without prior guidance, it struggles with high-dimensional exploration in complex, multi-stage tasks. To address these coupled generalization and exploration challenges, we present Dex-One2Many, a real-to-sim-to-real framework that learns a generalizable dexterous manipulation policy from a single human video. Our key insight is to abstract the video into sequential scene graphs that guide RL, enabling efficient exploration while preserving broad generalizability. The graphs serve as generative constraints for sampling diverse reset states and provide dense rewards for each stage. Because the graphs constrain relations rather than exact poses, these reset states cover object poses and grasps beyond the video, while initializing each stage from them with dense rewards keeps exploration short and guided. Trained entirely in simulation, Dex-One2Many transfers zero-shot to a real multi-fingered hand. Across five tool-use and manipulation tasks, Dex-One2Many exceeds baselines by 6.5% in seen configurations, while its robust generalization widens this gap to 71% in unseen scenarios.",
    "history": "Submission history From: Jusuk Lee [ view email ] [v1] Thu, 8 Oct 2026 17:59:58 UTC (25,950 KB)",
    "meta": {
      "citation_title": "Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration",
      "citation_author": "Lee, Seungjae",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12470",
      "citation_arxiv_id": "2610.12470",
      "citation_abstract": "While learning dexterous manipulation from a single human video offers a promising alternative to costly robot demonstrations, many recent methods predominantly imitate demonstrated motions. Such strict motion matching often limits generalization to initial object poses, goal poses, and grasps not shown in the video. Alternatively, discovering a policy via reinforcement learning (RL) allows for broad generalization, but without prior guidance, it struggles with high-dimensional exploration in complex, multi-stage tasks. To address these coupled generalization and exploration challenges, we present Dex-One2Many, a real-to-sim-to-real framework that learns a generalizable dexterous manipulation policy from a single human video. Our key insight is to abstract the video into sequential scene graphs that guide RL, enabling efficient exploration while preserving broad generalizability. The graphs serve as generative constraints for sampling diverse reset states and provide dense rewards for each stage. Because the graphs constrain relations rather than exact poses, these reset states cover object poses and grasps beyond the video, while initializing each stage from them with dense rewards keeps exploration short and guided. Trained entirely in simulation, Dex-One2Many transfers zero-shot to a real multi-fingered hand. Across five tool-use and manipulation tasks, Dex-One2Many exceeds baselines by 6.5% in seen configurations, while its robust generalization widens this gap to 71% in unseen scenarios."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-09T01:59:58+08:00"
  },
  {
    "id": "2610.12457v1",
    "title": "SpatialHarness: Test-Time Spatial Scaffolding for Fine Robotic Manipulation",
    "abstract": "Frontier multimodal foundation models (e.g., GPT-6 Astra) have recently shown strong potential for direct robotic control, yet their performance on fine manipulation remains limited. We argue that an important source of failure is not necessarily insufficient policy capability, but insufficient spatial observability, where task-critical spatial relationships may be poorly revealed by the existing physical camera setup. We introduce SpatialHarness, a test-time embodied harness that provides test-time spatial scaffolding for fine robotic manipulation without policy fine-tuning or changes to the physical sensing setup. SpatialHarness maintains an online simulated scene synchronized with real-world execution, identifies task-critical spatial relationships, and renders complementary virtual views that expose them to a frozen multimodal policy. To keep the simulated scene aligned during interaction, we develop interaction-aware scene synchronization that distinguishes static, held, and transition modes. We evaluate SpatialHarness on four real-robot manipulation tasks spanning precise geometric alignment, object-relative placement, and articulated-object interaction. Using the same frozen GPT-6 Astra policy, SpatialHarness substantially improves task success, including from 26.7% to 66.7% on plug insertion and from 0% to 100% on Tower of Hanoi. These results indicate that improving spatial observability at test time can unlock fine-manipulation capabilities already present in strong multimodal foundation models. Project website: this https URL .",
    "history": "Submission history From: Yue Yu [ view email ] [v1] Thu, 8 Oct 2026 17:59:20 UTC (5,886 KB)",
    "meta": {
      "citation_title": "SpatialHarness: Test-Time Spatial Scaffolding for Fine Robotic Manipulation",
      "citation_author": "Chen, Jingjing",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12457",
      "citation_arxiv_id": "2610.12457",
      "citation_abstract": "Frontier multimodal foundation models (e.g., GPT-6 Astra) have recently shown strong potential for direct robotic control, yet their performance on fine manipulation remains limited. We argue that an important source of failure is not necessarily insufficient policy capability, but insufficient spatial observability, where task-critical spatial relationships may be poorly revealed by the existing physical camera setup. We introduce SpatialHarness, a test-time embodied harness that provides test-time spatial scaffolding for fine robotic manipulation without policy fine-tuning or changes to the physical sensing setup. SpatialHarness maintains an online simulated scene synchronized with real-world execution, identifies task-critical spatial relationships, and renders complementary virtual views that expose them to a frozen multimodal policy. To keep the simulated scene aligned during interaction, we develop interaction-aware scene synchronization that distinguishes static, held, and transition modes. We evaluate SpatialHarness on four real-robot manipulation tasks spanning precise geometric alignment, object-relative placement, and articulated-object interaction. Using the same frozen GPT-6 Astra policy, SpatialHarness substantially improves task success, including from 26.7% to 66.7% on plug insertion and from 0% to 100% on Tower of Hanoi. These results indicate that improving spatial observability at test time can unlock fine-manipulation capabilities already present in strong multimodal foundation models. Project website: https://emilia113.github.io/SpatialHarness/."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-09T01:59:20+08:00"
  },
  {
    "id": "2610.12435v1",
    "title": "VioLA: Learning Generalist Humanoid Control Policies from Human Data",
    "abstract": "Teaching a humanoid to follow instructions with its whole body runs into two obstacles. Its action space is large and tightly coupled: legs, arms, and fingers must move together while the robot keeps its balance, which makes joint-level actions hard to learn. And humanoid demonstrations are scarce, so current humanoid generalist policies do not follow new instructions out of the box and are fine-tuned on teleoperated demonstrations of each task before deployment. Human demonstrations exist in far larger numbers, but a person's motion is not a robot command. We remove both obstacles by changing what the generalist policy predicts. We introduce VioLA, a generalist humanoid policy that predicts body and hand motion latents instead of joint commands. A pretrained body- and hand-controller execute these latents on the robot. Their corresponding motion encoders map human and robot motion into the same latent spaces. A human recording is therefore labeled in the policy's action space, and the training demonstration pool contains 140.6 million frames, 93.2% of them human. As a result, VioLA follows locomotion instructions on the real robot zero-shot, without task-specific fine-tuning, reaching 100% success where GR00T N1.7 and $\\Psi_0$ reach 16.7% and 0%, respectively. It also reaches 88.6% manipulation success without task-specific fine-tuning. The same approach works across two VLA and one world-action model backbones. A generalist policy trained on human demonstrations alone performs locomotion tasks on the real robot zero-shot. Code and checkpoints will be released.",
    "history": "Submission history From: Mert Albaba [ view email ] [v1] Thu, 8 Oct 2026 17:57:27 UTC (6,440 KB)",
    "meta": {
      "citation_title": "VioLA: Learning Generalist Humanoid Control Policies from Human Data",
      "citation_author": "Riedmiller, Martin",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12435",
      "citation_arxiv_id": "2610.12435",
      "citation_abstract": "Teaching a humanoid to follow instructions with its whole body runs into two obstacles. Its action space is large and tightly coupled: legs, arms, and fingers must move together while the robot keeps its balance, which makes joint-level actions hard to learn. And humanoid demonstrations are scarce, so current humanoid generalist policies do not follow new instructions out of the box and are fine-tuned on teleoperated demonstrations of each task before deployment. Human demonstrations exist in far larger numbers, but a person's motion is not a robot command. We remove both obstacles by changing what the generalist policy predicts. We introduce VioLA, a generalist humanoid policy that predicts body and hand motion latents instead of joint commands. A pretrained body- and hand-controller execute these latents on the robot. Their corresponding motion encoders map human and robot motion into the same latent spaces. A human recording is therefore labeled in the policy's action space, and the training demonstration pool contains 140.6 million frames, 93.2% of them human. As a result, VioLA follows locomotion instructions on the real robot zero-shot, without task-specific fine-tuning, reaching 100% success where GR00T N1.7 and $\\Psi_0$ reach 16.7% and 0%, respectively. It also reaches 88.6% manipulation success without task-specific fine-tuning. The same approach works across two VLA and one world-action model backbones. A generalist policy trained on human demonstrations alone performs locomotion tasks on the real robot zero-shot. Code and checkpoints will be released."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-09T01:57:27+08:00"
  },
  {
    "id": "2610.12407v1",
    "title": "LeWAM: A JEPA World Action Model with Diffusion-Steering-Based MPC",
    "abstract": "World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant information which can complicate downstream predictions. We introduce LeWAM, a bidirectional transformer for forward, backward, inverse dynamics and policy prediction, on a decoder-free JEPA latent trained end-to-end through all four modes. We see the following benefits: 1) Alignment: linear probes read robot and object state from LeWAM's latent better than from a regular Le World Model (a forward-only JEPA world model), while the latent ignores visual distractors as well as LeWM does and far better than a reconstruction-based WAM. 2) Acting: Closed-loop evaluations of LeWAM match a regular flow-matching policy trained on the same encoder at matched size, while also providing a world model. 3) Planning: Sampling raw actions when planning with WAMs lets MPC exploit dynamics-model inaccuracies; planning in the noise space of the policy head instead improves the closed-loop performance of these WAMs.",
    "history": "Submission history From: Shashank Hegde [ view email ] [v1] Thu, 8 Oct 2026 17:46:03 UTC (2,366 KB)",
    "meta": {
      "citation_title": "LeWAM: A JEPA World Action Model with Diffusion-Steering-Based MPC",
      "citation_author": "Smolyanskiy, Nikolai",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12407",
      "citation_arxiv_id": "2610.12407",
      "citation_abstract": "World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant information which can complicate downstream predictions. We introduce LeWAM, a bidirectional transformer for forward, backward, inverse dynamics and policy prediction, on a decoder-free JEPA latent trained end-to-end through all four modes. We see the following benefits: 1) Alignment: linear probes read robot and object state from LeWAM's latent better than from a regular Le World Model (a forward-only JEPA world model), while the latent ignores visual distractors as well as LeWM does and far better than a reconstruction-based WAM. 2) Acting: Closed-loop evaluations of LeWAM match a regular flow-matching policy trained on the same encoder at matched size, while also providing a world model. 3) Planning: Sampling raw actions when planning with WAMs lets MPC exploit dynamics-model inaccuracies; planning in the noise space of the policy head instead improves the closed-loop performance of these WAMs."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-09T01:46:03+08:00"
  },
  {
    "id": "2610.12194v1",
    "title": "MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling",
    "abstract": "World modeling has emerged as an effective co-training objective for robot policies, giving rise to World Action Models (WAMs) that jointly predict actions and future states. However, most WAMs predict future states in the native representation space of pretrained visual backbones, resulting in high-dimensional targets with substantial training cost. We introduce MiniWAM, which instead predicts compact future representations learned from privileged current-future transitions. To construct these targets, we propose Predictive Representations via Inverse Spatiotemporal Modeling (PRISM), which combines inverse-dynamics supervision with feature reconstruction to emphasize control-relevant transition information while preserving useful future-state information. With the learned PRISM encoder frozen, MiniWAM is trained to jointly predict the resulting targets and robot actions from current observations. With 65$\\times$ fewer native future feature tokens, MiniWAM consistently outperforms native future-feature prediction with both DINOv3 and WAN2.1 VAE features, while achieving up to an 8$\\times$ speedup in world-action training. At 0.25B parameters, MiniWAM is already competitive with substantially larger WAMs on LIBERO, LIBERO-Plus, and RoboTwin 2.0 simulation benchmarks. Representation analyses further show that PRISM contributes behavioral structure beyond feature reconstruction alone. These results demonstrate that effective world-action modeling does not require predicting native visual futures, and that compact predictive representations provide a strong and substantially more efficient target for policy learning. The project page is available at: this https URL .",
    "history": "Submission history From: Jie Chen [ view email ] [v1] Thu, 8 Oct 2026 15:51:53 UTC (1,457 KB)",
    "meta": {
      "citation_title": "MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling",
      "citation_author": "Sartoretti, Guillaume",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12194",
      "citation_arxiv_id": "2610.12194",
      "citation_abstract": "World modeling has emerged as an effective co-training objective for robot policies, giving rise to World Action Models (WAMs) that jointly predict actions and future states. However, most WAMs predict future states in the native representation space of pretrained visual backbones, resulting in high-dimensional targets with substantial training cost. We introduce MiniWAM, which instead predicts compact future representations learned from privileged current-future transitions. To construct these targets, we propose Predictive Representations via Inverse Spatiotemporal Modeling (PRISM), which combines inverse-dynamics supervision with feature reconstruction to emphasize control-relevant transition information while preserving useful future-state information. With the learned PRISM encoder frozen, MiniWAM is trained to jointly predict the resulting targets and robot actions from current observations. With 65$\\times$ fewer native future feature tokens, MiniWAM consistently outperforms native future-feature prediction with both DINOv3 and WAN2.1 VAE features, while achieving up to an 8$\\times$ speedup in world-action training. At 0.25B parameters, MiniWAM is already competitive with substantially larger WAMs on LIBERO, LIBERO-Plus, and RoboTwin 2.0 simulation benchmarks. Representation analyses further show that PRISM contributes behavioral structure beyond feature reconstruction alone. These results demonstrate that effective world-action modeling does not require predicting native visual futures, and that compact predictive representations provide a strong and substantially more efficient target for policy learning. The project page is available at: https://j1dan.github.io/MiniWAM."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-08T23:51:53+08:00"
  },
  {
    "id": "2610.12185v1",
    "title": "RESETTLE: Robotic Recovery through Disagreement-Triggered Retrieval and Efficient Corrective Control",
    "abstract": "Reliable robotic manipulation requires timely intervention to correct emerging deviations and restore progress after execution errors. However, recovery methods based on repeated vision-language reasoning or iterative online optimization can incur substantial latency, delaying intervention. To address these challenges, we introduce RESETTLE(Robotic rEcovery through diSagrEement-Triggered reTrievaL and Efficient Corrective Control), a model-agnostic framework that provides computationally efficient recovery at the action-execution interface of frozen robot policies. RESETTLE triggers recovery when two action proposals independently sampled under identical conditioning persistently disagree. It retrieves a same-task demonstration reference using an adapted V-JEPA encoder and combines a state-servo prior with a guarded visual residual to execute one corrective action without online trajectory optimization or additional vision-language reasoning, then returns control to the base policy. Across six base policies in simulation, RESETTLE achieves up to 8.70%, 6.28%, and 6.83% absolute success-rate gains on LIBERO-Plus, Meta-World, and RoboCasa Tabletop, respectively, with further improvements on four real-world tasks using two policies. In QwenPI-based comparisons, its monitoring-and-recovery computation latency is 74.04%--93.57% lower than VoLoAgent's monitoring-and-planning latency for grasp and place tool calls. It also raises Harness VLA's LIBERO-Pro Swap success from 42% to 50%, demonstrating compatibility with high-level agentic planning. Code available at: this https URL",
    "history": "Submission history From: Yuxin Chen [ view email ] [v1] Thu, 8 Oct 2026 15:49:03 UTC (8,051 KB)",
    "meta": {
      "citation_title": "RESETTLE: Robotic Recovery through Disagreement-Triggered Retrieval and Efficient Corrective Control",
      "citation_author": "Jia, Jiaya",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12185",
      "citation_arxiv_id": "2610.12185",
      "citation_abstract": "Reliable robotic manipulation requires timely intervention to correct emerging deviations and restore progress after execution errors. However, recovery methods based on repeated vision-language reasoning or iterative online optimization can incur substantial latency, delaying intervention. To address these challenges, we introduce RESETTLE(Robotic rEcovery through diSagrEement-Triggered reTrievaL and Efficient Corrective Control), a model-agnostic framework that provides computationally efficient recovery at the action-execution interface of frozen robot policies. RESETTLE triggers recovery when two action proposals independently sampled under identical conditioning persistently disagree. It retrieves a same-task demonstration reference using an adapted V-JEPA encoder and combines a state-servo prior with a guarded visual residual to execute one corrective action without online trajectory optimization or additional vision-language reasoning, then returns control to the base policy. Across six base policies in simulation, RESETTLE achieves up to 8.70%, 6.28%, and 6.83% absolute success-rate gains on LIBERO-Plus, Meta-World, and RoboCasa Tabletop, respectively, with further improvements on four real-world tasks using two policies. In QwenPI-based comparisons, its monitoring-and-recovery computation latency is 74.04%--93.57% lower than VoLoAgent's monitoring-and-planning latency for grasp and place tool calls. It also raises Harness VLA's LIBERO-Pro Swap success from 42% to 50%, demonstrating compatibility with high-level agentic planning. Code available at: https://github.com/JIA-Lab-research/RESETTLE"
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-08T23:49:03+08:00"
  },
  {
    "id": "2610.12089v1",
    "title": "ManiUnit: A Manipulation Skill Dataset and Benchmark for Long-Horizon Tasks",
    "abstract": "Long-horizon mobile manipulation requires a robot to navigate multi-room environments and execute a sequence of manipulation skills under a single natural language instruction. Learning and evaluating these skills present three challenges: similar observations under a fixed task instruction may make skill selection ambiguous; even when a preceding skill succeeds, the robot state inherited by the next skill may deviate from its demonstrated starting states and affect execution; and task-level metrics hinder skill-specific diagnosis, while early failures leave later skills untested. We therefore introduce ManiUnit, a manipulation skill dataset and benchmark built from 50 BEHAVIOR-1K activities. Its dataset contains 137,899 segments across 21 skill types and 417 subtasks, and its benchmark contains 1,260 test instances. Correspondingly, ManiUnit pairs each segment with an explicit subtask instruction; measures sensitivity to perturbations of the robot's starting base position or joint configuration; and restores intermediate simulator states and defines local success conditions so that each skill can be evaluated without executing preceding stages. Evaluations of representative vision-language-action (VLA) policies show that similar aggregate scores can hide substantial per-skill differences. The tested starting-state perturbations also degrade execution: on the full benchmark, joint perturbations reduce success rates by approximately 56% relative to those from demonstrated starting states. On two long-horizon activities, a skill policy trained on ManiUnit segments achieves 78.7% local manipulation success, compared with 49.3% for a task policy trained on complete demonstrations. The trained skills further support complete-task execution on these activities, as coordinating the task and skill policies through a planner raises full-task success from 4.0% to 18.0%.",
    "history": "Submission history From: Guoting Wei [ view email ] [v1] Thu, 8 Oct 2026 15:00:04 UTC (8,655 KB)",
    "meta": {
      "citation_title": "ManiUnit: A Manipulation Skill Dataset and Benchmark for Long-Horizon Tasks",
      "citation_author": "Xiao, Rong",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12089",
      "citation_arxiv_id": "2610.12089",
      "citation_abstract": "Long-horizon mobile manipulation requires a robot to navigate multi-room environments and execute a sequence of manipulation skills under a single natural language instruction. Learning and evaluating these skills present three challenges: similar observations under a fixed task instruction may make skill selection ambiguous; even when a preceding skill succeeds, the robot state inherited by the next skill may deviate from its demonstrated starting states and affect execution; and task-level metrics hinder skill-specific diagnosis, while early failures leave later skills untested. We therefore introduce ManiUnit, a manipulation skill dataset and benchmark built from 50 BEHAVIOR-1K activities. Its dataset contains 137,899 segments across 21 skill types and 417 subtasks, and its benchmark contains 1,260 test instances. Correspondingly, ManiUnit pairs each segment with an explicit subtask instruction; measures sensitivity to perturbations of the robot's starting base position or joint configuration; and restores intermediate simulator states and defines local success conditions so that each skill can be evaluated without executing preceding stages. Evaluations of representative vision-language-action (VLA) policies show that similar aggregate scores can hide substantial per-skill differences. The tested starting-state perturbations also degrade execution: on the full benchmark, joint perturbations reduce success rates by approximately 56% relative to those from demonstrated starting states. On two long-horizon activities, a skill policy trained on ManiUnit segments achieves 78.7% local manipulation success, compared with 49.3% for a task policy trained on complete demonstrations. The trained skills further support complete-task execution on these activities, as coordinating the task and skill policies through a planner raises full-task success from 4.0% to 18.0%."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-08T23:00:04+08:00"
  },
  {
    "id": "2610.12007v1",
    "title": "REACT: Rolling Denoising and Dual Decoupling for Reactive Robot Control with VLA Models",
    "abstract": "Flow-based vision-language-action (VLA) models generate action chunks for temporally coherent robot motion, but chunked control creates a fundamental closed-loop trade-off: long chunks provide smooth execution, whereas frequent replanning improves reactivity at the cost of action discontinuities. We introduce REACT, a rolling-denoising framework that makes flow-based VLAs more reactive while preserving long-horizon context. Instead of regenerating entire action chunks from scratch, REACT maintains a persistent action buffer with staggered flow timesteps. At each control step, the full horizon is denoised using the latest observation, the cleanest action block is executed, partially refined future blocks are shifted forward, and fresh noise is appended to the tail. As a result, each executed action block is refined across multiple recent observations before deployment. To support real-time control, we further introduce dual decoupling, which separates sensing, VLM encoding, DiT denoising, and action execution, enabling high-frequency observation updates and action streaming under practical compute constraints. Across the RoboTwin 2.0 simulation benchmark and real-world tasks spanning bimanual manipulation and dynamic control on multiple robot platforms, REACT improves task success and reduces reaction latency while producing smoother trajectories than frequent-replanning and asynchronous baselines.",
    "history": "Submission history From: Houlong Xiong [ view email ] [v1] Thu, 8 Oct 2026 14:09:46 UTC (16,529 KB)",
    "meta": {
      "citation_title": "REACT: Rolling Denoising and Dual Decoupling for Reactive Robot Control with VLA Models",
      "citation_author": "Zhu, Qian",
      "citation_date": "2026/10/08",
      "citation_online_date": "2026/10/08",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.12007",
      "citation_arxiv_id": "2610.12007",
      "citation_abstract": "Flow-based vision-language-action (VLA) models generate action chunks for temporally coherent robot motion, but chunked control creates a fundamental closed-loop trade-off: long chunks provide smooth execution, whereas frequent replanning improves reactivity at the cost of action discontinuities. We introduce REACT, a rolling-denoising framework that makes flow-based VLAs more reactive while preserving long-horizon context. Instead of regenerating entire action chunks from scratch, REACT maintains a persistent action buffer with staggered flow timesteps. At each control step, the full horizon is denoised using the latest observation, the cleanest action block is executed, partially refined future blocks are shifted forward, and fresh noise is appended to the tail. As a result, each executed action block is refined across multiple recent observations before deployment. To support real-time control, we further introduce dual decoupling, which separates sensing, VLM encoding, DiT denoising, and action execution, enabling high-frequency observation updates and action streaming under practical compute constraints. Across the RoboTwin 2.0 simulation benchmark and real-world tasks spanning bimanual manipulation and dynamic control on multiple robot platforms, REACT improves task success and reduces reaction latency while producing smoother trajectories than frequent-replanning and asynchronous baselines."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-08T22:09:46+08:00"
  },
  {
    "id": "2610.10409v1",
    "title": "RobotWorld: Benchmarking Multimodal Agents for Robot Use Across Diverse Tasks and Embodiments",
    "abstract": "General-purpose agents increasingly write code, use tools, and complete complex digital tasks, raising the question of how far these capabilities carry into the physical world. To investigate this, we introduce RobotWorld, a challenging simulation testbed for robot use: turning instructions and observations into physical task execution through robot interfaces. Its 84 tasks span manipulation, mobile manipulation, locomotion, driving, and aerial control, with explicit interaction budgets and executable success checks. By analysing task outcomes alongside execution traces, we identify both the capabilities that transfer and the gaps that prevent reliable completion. Furthermore, we find that current agents can construct sophisticated perception and control workflows, including image segmentation, camera calibration, spatial estimation, and dynamics-based computation. These capabilities, however, do not consistently compose into successful behaviour: agents lose task-relevant object states despite reaching commanded poses, fail to correct ineffective actions, recover too late, or mistake unfinished tasks for completion. This uneven transfer also differs across models: Astra succeeds more often on spatial and constrained-contact goals, whereas Opus 5.5 succeeds more often on continuous-balance and timed-interaction goals. By linking these outcomes to execution behaviour, RobotWorld provides both a rigorous proving ground and an empirical account of the remaining capability gaps, thereby establishing concrete targets for training and designing more reliable physical-world agents.",
    "history": "Submission history From: Zhiqin Yang [ view email ] [v1] Wed, 7 Oct 2026 16:55:24 UTC (11,228 KB)",
    "meta": {
      "citation_title": "RobotWorld: Benchmarking Multimodal Agents for Robot Use Across Diverse Tasks and Embodiments",
      "citation_author": "Guo, Yike",
      "citation_date": "2026/10/07",
      "citation_online_date": "2026/10/07",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.10409",
      "citation_arxiv_id": "2610.10409",
      "citation_abstract": "General-purpose agents increasingly write code, use tools, and complete complex digital tasks, raising the question of how far these capabilities carry into the physical world. To investigate this, we introduce RobotWorld, a challenging simulation testbed for robot use: turning instructions and observations into physical task execution through robot interfaces. Its 84 tasks span manipulation, mobile manipulation, locomotion, driving, and aerial control, with explicit interaction budgets and executable success checks. By analysing task outcomes alongside execution traces, we identify both the capabilities that transfer and the gaps that prevent reliable completion. Furthermore, we find that current agents can construct sophisticated perception and control workflows, including image segmentation, camera calibration, spatial estimation, and dynamics-based computation. These capabilities, however, do not consistently compose into successful behaviour: agents lose task-relevant object states despite reaching commanded poses, fail to correct ineffective actions, recover too late, or mistake unfinished tasks for completion. This uneven transfer also differs across models: Astra succeeds more often on spatial and constrained-contact goals, whereas Opus 5.5 succeeds more often on continuous-balance and timed-interaction goals. By linking these outcomes to execution behaviour, RobotWorld provides both a rigorous proving ground and an empirical account of the remaining capability gaps, thereby establishing concrete targets for training and designing more reliable physical-world agents."
    },
    "groups": [
      "w5"
    ],
    "event_at": "2026-10-08T00:55:24+08:00"
  },
  {
    "id": "2607.15330v1",
    "title": "Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectories",
    "abstract": "We present Xiaomi-Robotics-1, a foundational vision-language-action (VLA) model capable of (1) following diverse language instructions to perform a wide range of mobile manipulation tasks in unseen environments out-of-the-box, and (2) efficiently adapting to novel downstream tasks with minimal fine-tuning data. We propose a two-stage training recipe consisting of pre-training and post-training. During pre-training, we imbue the model with broad and generalizable action-generation capabilities by training on over 100k hours of real-world manipulation trajectories collected via UMI devices. Crucially, we develop a scalable auto-labeling pipeline that annotates trajectory clips with natural languages describing scene state transitions, providing rich and precise conditioning for action learning. During post-training, we aim to align these capabilities with robot embodiments and imperative instructions that humans naturally use to prompt robots. Extensive experiments demonstrate strong scaling behavior. Xiaomi-Robotics-1 consistently improves with increased data scales and model sizes during pre-training. This scaling behavior directly transfers to post-training, where a stronger pre-training model yields better out-of-the-box real-robot performance in unseen environments. Furthermore, Xiaomi-Robotics-1 serves as a strong robot foundation policy that can be efficiently fine-tuned on complex, dexterous tasks with high data efficiency. Across multiple simulation benchmarks, Xiaomi-Robotics-1 outperforms state-of-the-art methods. Notably, it establishes a new state-of-the-art with a 57.6% success rate on RoboCasa365, surpassing the previous best of 46.6%. Furthermore, it achieves an average score of 20.07 on RoboDojo, significantly outperforming the prior state-of-the-art (13.07). Code and model checkpoints will be released. Project page: this https URL",
    "history": "Submission history From: Yifei Su [ view email ] [v1] Thu, 16 Jul 2026 16:02:25 UTC (15,643 KB) [v2] Wed, 22 Jul 2026 11:38:14 UTC (15,643 KB)",
    "meta": {
      "citation_title": "Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectories",
      "citation_author": "Zhou, Quanyun",
      "citation_date": "2026/07/16",
      "citation_online_date": "2026/07/16",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.15330",
      "citation_arxiv_id": "2607.15330",
      "citation_abstract": "We present Xiaomi-Robotics-1, a foundational vision-language-action (VLA) model capable of (1) following diverse language instructions to perform a wide range of mobile manipulation tasks in unseen environments out-of-the-box, and (2) efficiently adapting to novel downstream tasks with minimal fine-tuning data. We propose a two-stage training recipe consisting of pre-training and post-training. During pre-training, we imbue the model with broad and generalizable action-generation capabilities by training on over 100k hours of real-world manipulation trajectories collected via UMI devices. Crucially, we develop a scalable auto-labeling pipeline that annotates trajectory clips with natural languages describing scene state transitions, providing rich and precise conditioning for action learning. During post-training, we aim to align these capabilities with robot embodiments and imperative instructions that humans naturally use to prompt robots. Extensive experiments demonstrate strong scaling behavior. Xiaomi-Robotics-1 consistently improves with increased data scales and model sizes during pre-training. This scaling behavior directly transfers to post-training, where a stronger pre-training model yields better out-of-the-box real-robot performance in unseen environments. Furthermore, Xiaomi-Robotics-1 serves as a strong robot foundation policy that can be efficiently fine-tuned on complex, dexterous tasks with high data efficiency. Across multiple simulation benchmarks, Xiaomi-Robotics-1 outperforms state-of-the-art methods. Notably, it establishes a new state-of-the-art with a 57.6% success rate on RoboCasa365, surpassing the previous best of 46.6%. Furthermore, it achieves an average score of 20.07 on RoboDojo, significantly outperforming the prior state-of-the-art (13.07). Code and model checkpoints will be released. Project page: https://robotics.xiaomi.com/xiaomi-robotics-1.html"
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-17T00:02:25+08:00"
  },
  {
    "id": "2607.16187v1",
    "title": "Handroid: Bridging Dexterous Hand and Humanoid",
    "abstract": "Dexterous hands and humanoid robots are typically developed as distinct embodiments: the former enable contact-rich manipulation at the object scale, whereas the latter provide mobility and whole-body interaction in human-centered environments. We introduce \\textbf{Handroid}, a desktop-scale dual-embodiment robot that integrates both capabilities within a single reconfigurable platform. Handroid reuses one 27-DoF electromechanical body as either a dexterous hand or a desktop humanoid, measuring 0.33 m in height and 2.05 kg in weight. In the dexterous hand embodiment, 20 DoFs form an anthropomorphic hand closely matching the kinematic structure of the human hand. In the humanoid embodiment, the same articulated modules are reconfigured into a humanoid with a head, arms, and legs, including a 12-DoF lower-limb structure for locomotion and whole-body motion. Handroid further provides a unified control and learning framework supporting hand teleoperation, dexterous grasping, in-hand manipulation, humanoid locomotion, gait generation, and interactive motion authoring. We validate the platform through real-world dexterous manipulation, reinforcement-learning-based locomotion, keyframe motion deployment, and a long-horizon task involving embodiment reconfiguration, locomotion, docking, and dexterous pick-and-place. These results position Handroid as a compact and reproducible platform for advancing morphology-reconfigurable robotics and cross-embodiment robot learning.",
    "history": "Submission history From: Chenyang Ma [ view email ] [v1] Fri, 17 Jul 2026 17:59:05 UTC (18,328 KB)",
    "meta": {
      "citation_title": "Handroid: Bridging Dexterous Hand and Humanoid",
      "citation_author": "Ding, Mingyu",
      "citation_date": "2026/07/17",
      "citation_online_date": "2026/07/17",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.16187",
      "citation_arxiv_id": "2607.16187",
      "citation_abstract": "Dexterous hands and humanoid robots are typically developed as distinct embodiments: the former enable contact-rich manipulation at the object scale, whereas the latter provide mobility and whole-body interaction in human-centered environments. We introduce \\textbf{Handroid}, a desktop-scale dual-embodiment robot that integrates both capabilities within a single reconfigurable platform. Handroid reuses one 27-DoF electromechanical body as either a dexterous hand or a desktop humanoid, measuring 0.33 m in height and 2.05 kg in weight. In the dexterous hand embodiment, 20 DoFs form an anthropomorphic hand closely matching the kinematic structure of the human hand. In the humanoid embodiment, the same articulated modules are reconfigured into a humanoid with a head, arms, and legs, including a 12-DoF lower-limb structure for locomotion and whole-body motion. Handroid further provides a unified control and learning framework supporting hand teleoperation, dexterous grasping, in-hand manipulation, humanoid locomotion, gait generation, and interactive motion authoring. We validate the platform through real-world dexterous manipulation, reinforcement-learning-based locomotion, keyframe motion deployment, and a long-horizon task involving embodiment reconfiguration, locomotion, docking, and dexterous pick-and-place. These results position Handroid as a compact and reproducible platform for advancing morphology-reconfigurable robotics and cross-embodiment robot learning."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-18T01:59:05+08:00"
  },
  {
    "id": "2607.15982v1",
    "title": "Data and Learning Where it Matters for Contact-Rich Manipulation",
    "abstract": "Learned policies trained end-to-end on large datasets often remain brittle in high-precision tasks and struggle with generalization. We find that these limitations largely stem from a lack of structure and focus in data collection. Our key insight is to leverage dense data collection only for the critical segment of contact-rich tasks and to rely on traditional planning during simple free-space motion. We propose an automated data-collection scheme in combination with offline deep reinforcement learning for the critical segment of the task, eliminating reliance on a teleoperator's skill and on online policy updates. Across four challenging real-world tasks, using only 2 to 2.5 hours of autonomous data collection, we achieve an average success rate of 96%, compared to the strongest baseline at 55%. Notably, performance remains high in out-of-distribution scenarios where end-to-end approaches struggle. Our results pave the way for targeted data collection for contact-rich tasks and for high success rates in precision applications.",
    "history": "Submission history From: Oliver Hausdörfer [ view email ] [v1] Fri, 17 Jul 2026 14:18:20 UTC (28,385 KB)",
    "meta": {
      "citation_title": "Data and Learning Where it Matters for Contact-Rich Manipulation",
      "citation_author": "Schoellig, Angela P.",
      "citation_date": "2026/07/17",
      "citation_online_date": "2026/07/17",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.15982",
      "citation_arxiv_id": "2607.15982",
      "citation_abstract": "Learned policies trained end-to-end on large datasets often remain brittle in high-precision tasks and struggle with generalization. We find that these limitations largely stem from a lack of structure and focus in data collection. Our key insight is to leverage dense data collection only for the critical segment of contact-rich tasks and to rely on traditional planning during simple free-space motion. We propose an automated data-collection scheme in combination with offline deep reinforcement learning for the critical segment of the task, eliminating reliance on a teleoperator's skill and on online policy updates. Across four challenging real-world tasks, using only 2 to 2.5 hours of autonomous data collection, we achieve an average success rate of 96%, compared to the strongest baseline at 55%. Notably, performance remains high in out-of-distribution scenarios where end-to-end approaches struggle. Our results pave the way for targeted data collection for contact-rich tasks and for high success rates in precision applications."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-17T22:18:20+08:00"
  },
  {
    "id": "2607.14183v1",
    "title": "Open-AoE: An Open Egocentric Manipulation Dataset and Toolchain for Embodied Learning",
    "abstract": "Egocentric videos of human manipulation provide scalable supervision for embodied intelligence, yet existing resources rarely combine low-cost continuous capture, manipulation-level structured annotations, and reusable tools for robot learning. We present Open-AoE, an open, community-oriented egocentric manipulation dataset and toolchain spanning the full pipeline from smartphone capture to model training. Its first release contains approximately 2,000 hours of manipulation video collected in natural environments by 500+ contributors using 400+ smartphones. The dataset provides text annotations, MANO-based hand poses, camera trajectories, and temporally localized atomic actions. Open-AoE further includes a data processing pipeline that transforms raw recordings into structured samples through temporal action segmentation, semantic annotation, hand reconstruction, and camera trajectory reconstruction. Meanwhile, we provide a separate downstream toolchain supports visualization, cross-embodiment retargeting, model-specific data conversion, and training recipes for VLA policies, WAMs, and World Models. By integrating scalable capture, structured processing, and downstream adaptation, Open-AoE reduces the barriers to both data contribution and reuse, providing practical open infrastructure for embodied model training, human-to-robot transfer, and world modeling.",
    "history": "Submission history From: Bowen Yang [ view email ] [v1] Wed, 15 Jul 2026 14:49:49 UTC (9,172 KB) [v2] Sat, 18 Jul 2026 10:02:02 UTC (9,172 KB)",
    "meta": {
      "citation_title": "Open-AoE: An Open Egocentric Manipulation Dataset and Toolchain for Embodied Learning",
      "citation_author": "Li, Zhe",
      "citation_date": "2026/07/15",
      "citation_online_date": "2026/07/15",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.14183",
      "citation_arxiv_id": "2607.14183",
      "citation_abstract": "Egocentric videos of human manipulation provide scalable supervision for embodied intelligence, yet existing resources rarely combine low-cost continuous capture, manipulation-level structured annotations, and reusable tools for robot learning. We present Open-AoE, an open, community-oriented egocentric manipulation dataset and toolchain spanning the full pipeline from smartphone capture to model training. Its first release contains approximately 2,000 hours of manipulation video collected in natural environments by 500+ contributors using 400+ smartphones. The dataset provides text annotations, MANO-based hand poses, camera trajectories, and temporally localized atomic actions. Open-AoE further includes a data processing pipeline that transforms raw recordings into structured samples through temporal action segmentation, semantic annotation, hand reconstruction, and camera trajectory reconstruction. Meanwhile, we provide a separate downstream toolchain supports visualization, cross-embodiment retargeting, model-specific data conversion, and training recipes for VLA policies, WAMs, and World Models. By integrating scalable capture, structured processing, and downstream adaptation, Open-AoE reduces the barriers to both data contribution and reuse, providing practical open infrastructure for embodied model training, human-to-robot transfer, and world modeling."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-15T22:49:49+08:00"
  },
  {
    "id": "2607.10706v1",
    "title": "Action Map Policy: Learning 3D Closed-loop Manipulation via Pixel Classification",
    "abstract": "The action space poses a major challenge in robot learning, since it is often high-dimensional, can span long time horizons, and frequently admits multi-modal optimal solutions. A good choice of action representation and loss function can help to address these concerns, but there are often trade offs. We propose Action Map Policy (AMP), which casts 3D closed-loop manipulation policy learning as a classification problem in image space. While classification has been an effective formulation in generative language models, applying it to robot action learning is difficult because naively discretizing high-dimensional continuous actions explodes the token vocabulary. Our key idea is to project 3D actions onto the camera image planes and treat each pixel location as a discrete class, thus controlling dimensionality while retaining multi-modality. This method supports millimeter-level precision for high-dimensional actions without requiring a prohibitively large vocabulary, while preserving fine-grained pixel-wise visual signals. Furthermore, it can predict the entire action chunk in a single forward pass, avoiding complex noise scheduling and iterative denoising while achieving substantially faster inference than diffusion policies. Experiments on various manipulation tasks show that AMP outperforms strong baselines, achieving higher success rates, faster inference, and enhanced spatial reasoning.",
    "history": "Submission history From: Haojie Huang [ view email ] [v1] Sun, 12 Jul 2026 10:58:06 UTC (18,202 KB)",
    "meta": {
      "citation_title": "Action Map Policy: Learning 3D Closed-loop Manipulation via Pixel Classification",
      "citation_author": "Walters, Robin",
      "citation_date": "2026/07/12",
      "citation_online_date": "2026/07/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.10706",
      "citation_arxiv_id": "2607.10706",
      "citation_abstract": "The action space poses a major challenge in robot learning, since it is often high-dimensional, can span long time horizons, and frequently admits multi-modal optimal solutions. A good choice of action representation and loss function can help to address these concerns, but there are often trade offs. We propose Action Map Policy (AMP), which casts 3D closed-loop manipulation policy learning as a classification problem in image space. While classification has been an effective formulation in generative language models, applying it to robot action learning is difficult because naively discretizing high-dimensional continuous actions explodes the token vocabulary. Our key idea is to project 3D actions onto the camera image planes and treat each pixel location as a discrete class, thus controlling dimensionality while retaining multi-modality. This method supports millimeter-level precision for high-dimensional actions without requiring a prohibitively large vocabulary, while preserving fine-grained pixel-wise visual signals. Furthermore, it can predict the entire action chunk in a single forward pass, avoiding complex noise scheduling and iterative denoising while achieving substantially faster inference than diffusion policies. Experiments on various manipulation tasks show that AMP outperforms strong baselines, achieving higher success rates, faster inference, and enhanced spatial reasoning."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-12T18:58:06+08:00"
  },
  {
    "id": "2607.25895v1",
    "title": "HiFi-UMI: Learning Deployable Manipulation Policies from High-Fidelity UMI Data Alone",
    "abstract": "Learning deployable manipulation policies is bottlenecked by the scarcity of data that is both high-fidelity and scalable. Real-robot teleoperation is accurate but costly to scale; robot-free UMI capture scales readily, and current practice uses the resulting data mainly for pre-training, adding a small real-robot \"anchor\" at post-training. We ask whether raising the fidelity of robot-free UMI data, rather than shrinking the real-robot fraction, can remove that anchor. We present HiFi-UMI, a portable UMI data-production system co-designed for trajectory accuracy, inter-gripper relative pose, synchronization, and field of view: head-mounted offline stereo-inertial SLAM, native rather than reconstructed relative pose, a shared microsecond GPIO trigger, and two wide-angle cameras per hand covering ~200 degrees. It reaches 3 mm workspace-local end-effector accuracy without external tracking infrastructure. Using this corpus, we demonstrate zero-robot post-training: a policy post-trained solely on HiFi-UMI demonstrations deploys directly on a real robot and matches in-domain teleoperation across three backbones spanning the vision-language-action and world-action-model families, with success-rate differences of -2.5, +3.1, and -0.6 percentage points on StarVLA-QwenPI, OpenPI-pi_0.5, and LingBot-VA; the strongest policy reaches 85% on a precision insertion task, even though the teleoperation baseline is collected in the evaluation scene and no HiFi-UMI trajectory is. Pre-training on 4,000 hours from the same corpus lowers action error on ten unseen tasks by 41% and, on StarVLA-QwenPI, raises real-robot success by a further 18.1 percentage points. We open-source HiFi-UMI-2K, 2,000 hours of microsecond-synchronized, ultra-wide-FoV demonstrations, each automatically reconstructed and validated through simulation replay, as a large-scale, high-fidelity resource for the robot-learning community.",
    "history": "Submission history From: Minglei Li [ view email ] [v1] Tue, 28 Jul 2026 15:52:02 UTC (20,656 KB)",
    "meta": {
      "citation_title": "HiFi-UMI: Learning Deployable Manipulation Policies from High-Fidelity UMI Data Alone",
      "citation_author": "Li, Xiaofei",
      "citation_date": "2026/07/28",
      "citation_online_date": "2026/07/28",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.25895",
      "citation_arxiv_id": "2607.25895",
      "citation_abstract": "Learning deployable manipulation policies is bottlenecked by the scarcity of data that is both high-fidelity and scalable. Real-robot teleoperation is accurate but costly to scale; robot-free UMI capture scales readily, and current practice uses the resulting data mainly for pre-training, adding a small real-robot \"anchor\" at post-training. We ask whether raising the fidelity of robot-free UMI data, rather than shrinking the real-robot fraction, can remove that anchor. We present HiFi-UMI, a portable UMI data-production system co-designed for trajectory accuracy, inter-gripper relative pose, synchronization, and field of view: head-mounted offline stereo-inertial SLAM, native rather than reconstructed relative pose, a shared microsecond GPIO trigger, and two wide-angle cameras per hand covering ~200 degrees. It reaches 3 mm workspace-local end-effector accuracy without external tracking infrastructure. Using this corpus, we demonstrate zero-robot post-training: a policy post-trained solely on HiFi-UMI demonstrations deploys directly on a real robot and matches in-domain teleoperation across three backbones spanning the vision-language-action and world-action-model families, with success-rate differences of -2.5, +3.1, and -0.6 percentage points on StarVLA-QwenPI, OpenPI-pi_0.5, and LingBot-VA; the strongest policy reaches 85% on a precision insertion task, even though the teleoperation baseline is collected in the evaluation scene and no HiFi-UMI trajectory is. Pre-training on 4,000 hours from the same corpus lowers action error on ten unseen tasks by 41% and, on StarVLA-QwenPI, raises real-robot success by a further 18.1 percentage points. We open-source HiFi-UMI-2K, 2,000 hours of microsecond-synchronized, ultra-wide-FoV demonstrations, each automatically reconstructed and validated through simulation replay, as a large-scale, high-fidelity resource for the robot-learning community."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-28T23:52:02+08:00"
  },
  {
    "id": "2607.18231v1",
    "title": "FM-VLA: Force-based Memory for Vision-Language-Action Models in Contact-Rich Manipulation",
    "abstract": "Vision-language-action (VLA) models have achieved impressive generalization in robotic manipulation, and recent memory-augmented VLAs have relaxed the Markovian assumption by conditioning on past images or language summaries. Vision-based memory approaches address this by conditioning on sampled past image frames, but they are computationally expensive and fundamentally limited when temporal events are visually ambiguous, e.g., pushing a button multiple times with small movements. We propose FM-VLA, a VLA model with force-based memory, enabling temporal context reasoning for non-Markovian, contact-rich manipulation. We encode force histories into compact force memory tokens with a variational autoencoder (VAE) pretrained with force time series reconstruction. By projecting force latent representations and short state history as additional conditioning tokens to the action expert module, we enable VLAs to leverage accumulated contact event history to guide manipulation. We evaluate FM-VLA on three memory-dependent tasks, including finding a hidden block, pressing a button, and wiping a dish for a specific number of times. Our lightweight force memory achieves over 80% success rate with minimal inference overhead, significantly outperforming baseline approaches. Project page: this https URL",
    "history": "Submission history From: Ruicheng Li [ view email ] [v1] Mon, 20 Jul 2026 17:58:31 UTC (1,919 KB)",
    "meta": {
      "citation_title": "FM-VLA: Force-based Memory for Vision-Language-Action Models in Contact-Rich Manipulation",
      "citation_author": "Guo, Baining",
      "citation_date": "2026/07/20",
      "citation_online_date": "2026/07/20",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.18231",
      "citation_arxiv_id": "2607.18231",
      "citation_abstract": "Vision-language-action (VLA) models have achieved impressive generalization in robotic manipulation, and recent memory-augmented VLAs have relaxed the Markovian assumption by conditioning on past images or language summaries. Vision-based memory approaches address this by conditioning on sampled past image frames, but they are computationally expensive and fundamentally limited when temporal events are visually ambiguous, e.g., pushing a button multiple times with small movements. We propose FM-VLA, a VLA model with force-based memory, enabling temporal context reasoning for non-Markovian, contact-rich manipulation. We encode force histories into compact force memory tokens with a variational autoencoder (VAE) pretrained with force time series reconstruction. By projecting force latent representations and short state history as additional conditioning tokens to the action expert module, we enable VLAs to leverage accumulated contact event history to guide manipulation. We evaluate FM-VLA on three memory-dependent tasks, including finding a hidden block, pressing a button, and wiping a dish for a specific number of times. Our lightweight force memory achieves over 80% success rate with minimal inference overhead, significantly outperforming baseline approaches. Project page: https://qft-333.github.io/FM-VLA-Page/"
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-21T01:58:31+08:00"
  },
  {
    "id": "2607.21588v1",
    "title": "AXIS: A Growable Community-Driven Data Engine for Scalable Robot Manipulation",
    "abstract": "Learning effective robot manipulation policies requires diverse, high-quality demonstrations, yet existing data pipelines are often difficult to scale because they rely on specialized hardware, centralized operators, or fixed task suites. We present AXIS, a growable community-driven data engine and benchmark for scalable robot learning, which enables browser-based teleoperation for large-scale demonstration collection, automatically generates and validates new manipulation tasks, and transforms community-collected demonstrations into training-ready data through automated success checking, quality filtering, trajectory smoothing, and visual and physics-based augmentation. The AXIS dataset currently contains 207 diverse tasks and 50K+ trajectories. Meanwhile, AXIS organizes data into task snapshots and evaluates policies with a systematic held-out protocol. We compare vision-language-action (VLA) policies under a unified AXIS evaluation suite and analyze scaling behavior across different data volumes. Continual pretraining on AXIS substantially improves the overall success rate of $\\pi_{0.5}$ by 5.8%, outperforms the model pretrained on RoboCasa365 by 37.3%, and exhibits consistent scaling with increasing data volume, with the largest gains observed under layout, sensor-noise, and camera perturbations.",
    "history": "Submission history From: Dihong Huang [ view email ] [v1] Thu, 23 Jul 2026 17:58:08 UTC (45,419 KB)",
    "meta": {
      "citation_title": "AXIS: A Growable Community-Driven Data Engine for Scalable Robot Manipulation",
      "citation_author": "Li, Jiachen",
      "citation_date": "2026/07/23",
      "citation_online_date": "2026/07/23",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.21588",
      "citation_arxiv_id": "2607.21588",
      "citation_abstract": "Learning effective robot manipulation policies requires diverse, high-quality demonstrations, yet existing data pipelines are often difficult to scale because they rely on specialized hardware, centralized operators, or fixed task suites. We present AXIS, a growable community-driven data engine and benchmark for scalable robot learning, which enables browser-based teleoperation for large-scale demonstration collection, automatically generates and validates new manipulation tasks, and transforms community-collected demonstrations into training-ready data through automated success checking, quality filtering, trajectory smoothing, and visual and physics-based augmentation. The AXIS dataset currently contains 207 diverse tasks and 50K+ trajectories. Meanwhile, AXIS organizes data into task snapshots and evaluates policies with a systematic held-out protocol. We compare vision-language-action (VLA) policies under a unified AXIS evaluation suite and analyze scaling behavior across different data volumes. Continual pretraining on AXIS substantially improves the overall success rate of $\\pi_{0.5}$ by 5.8%, outperforms the model pretrained on RoboCasa365 by 37.3%, and exhibits consistent scaling with increasing data volume, with the largest gains observed under layout, sensor-noise, and camera perturbations."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-24T01:58:08+08:00"
  },
  {
    "id": "2607.23108v1",
    "title": "The Curse of Precision: A Data Scaling Law for High-Precision Robotic Manipulation",
    "abstract": "While scaling laws for imitation learning have primarily focused on generalization in open-world settings, the relationship between data and precision in closed-world tasks like robotic assembly remains largely unexplored. This paper systematically investigates this relationship and introduces a novel scaling law. We find that to achieve a fixed success rate, the required number of demonstrations $N$ grows super-exponentially as the target precision $P$ approaches a limit $c$. This relationship is accurately captured by the model $\\log(N) \\propto 1/(P-c)$. Crucially, we reveal that the limit precision $c$ is not a static physical constant of the task but an emergent property of the entire agent system, including its sensors and expert policy. Through experiments on canonical manipulation tasks, we validate this law and demonstrate that improving system components, such as adding a wrist camera or using a more effective expert, measurably lowers $c$, thus expanding the system's achievable precision. Our work provides a new theoretical framework for precision in robotics and a quantitative metric to evaluate system capabilities. Furthermore, these findings provide a practical methodology for guiding the development and debugging of high-precision manipulation systems.",
    "history": "Submission history From: Cuijie Xu [ view email ] [v1] Sat, 25 Jul 2026 08:41:52 UTC (3,893 KB)",
    "meta": {
      "citation_title": "The Curse of Precision: A Data Scaling Law for High-Precision Robotic Manipulation",
      "citation_author": "Yu, Jincheng",
      "citation_date": "2026/07/25",
      "citation_online_date": "2026/07/25",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.23108",
      "citation_arxiv_id": "2607.23108",
      "citation_abstract": "While scaling laws for imitation learning have primarily focused on generalization in open-world settings, the relationship between data and precision in closed-world tasks like robotic assembly remains largely unexplored. This paper systematically investigates this relationship and introduces a novel scaling law. We find that to achieve a fixed success rate, the required number of demonstrations $N$ grows super-exponentially as the target precision $P$ approaches a limit $c$. This relationship is accurately captured by the model $\\log(N) \\propto 1/(P-c)$. Crucially, we reveal that the limit precision $c$ is not a static physical constant of the task but an emergent property of the entire agent system, including its sensors and expert policy. Through experiments on canonical manipulation tasks, we validate this law and demonstrate that improving system components, such as adding a wrist camera or using a more effective expert, measurably lowers $c$, thus expanding the system's achievable precision. Our work provides a new theoretical framework for precision in robotics and a quantitative metric to evaluate system capabilities. Furthermore, these findings provide a practical methodology for guiding the development and debugging of high-precision manipulation systems."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-25T16:41:52+08:00"
  },
  {
    "id": "2607.26789v1",
    "title": "CheckVLA: Execution-Time Verification with Action-Conditioned World Model for Long-Horizon Mobile Manipulation",
    "abstract": "Vision-language-action (VLA) policies commonly execute long-horizon mobile manipulation through open-loop action chunks, issuing multiple actions without receiving new high-level visual input. A committed chunk therefore implies how observations should evolve, but accidental deviations can violate this expectation while the remaining actions continue to propagate the error: commit-time policy confidence cannot react to a deviation that occurs after dispatch, and observation-only anomaly scores lack an action-conditioned reference for separating expected effects from unexplained changes. We propose CheckVLA, which verifies execution with a separately trained, frozen action-conditioned world model. A conformally calibrated risk threshold bounds the episode-level probability of an unnecessary first intervention and determines when to intervene, its exceedance controls how strongly the rewritten suffix retains the superseded chunk, latency-aware hard prefixing restricts replacement to actions that remain deployable, and an event-driven keyframe bank preserves evidence of prior progress across repairs. On RoboCasa365, under a common training recipe and a matched invocation budget, CheckVLA attains a 36.1% average success rate against 27.6% for periodic replanning (+8.5 points). At a matched 5% episode-level false-alarm target, action conditioning raises timely recall to 77.9%, against 48.6% for an observation-only control and 37.9% for an action-shuffled control. These simulation results support action-conditioned verification as a way to restore feedback during chunked execution while keeping the repair consistent with inference latency.",
    "history": "Submission history From: Yushan Liu [ view email ] [v1] Wed, 29 Jul 2026 11:31:33 UTC (2,390 KB)",
    "meta": {
      "citation_title": "CheckVLA: Execution-Time Verification with Action-Conditioned World Model for Long-Horizon Mobile Manipulation",
      "citation_author": "Ding, Wenbo",
      "citation_date": "2026/07/29",
      "citation_online_date": "2026/07/29",
      "citation_pdf_url": "https://arxiv.org/pdf/2607.26789",
      "citation_arxiv_id": "2607.26789",
      "citation_abstract": "Vision-language-action (VLA) policies commonly execute long-horizon mobile manipulation through open-loop action chunks, issuing multiple actions without receiving new high-level visual input. A committed chunk therefore implies how observations should evolve, but accidental deviations can violate this expectation while the remaining actions continue to propagate the error: commit-time policy confidence cannot react to a deviation that occurs after dispatch, and observation-only anomaly scores lack an action-conditioned reference for separating expected effects from unexplained changes. We propose CheckVLA, which verifies execution with a separately trained, frozen action-conditioned world model. A conformally calibrated risk threshold bounds the episode-level probability of an unnecessary first intervention and determines when to intervene, its exceedance controls how strongly the rewritten suffix retains the superseded chunk, latency-aware hard prefixing restricts replacement to actions that remain deployable, and an event-driven keyframe bank preserves evidence of prior progress across repairs. On RoboCasa365, under a common training recipe and a matched invocation budget, CheckVLA attains a 36.1% average success rate against 27.6% for periodic replanning (+8.5 points). At a matched 5% episode-level false-alarm target, action conditioning raises timely recall to 77.9%, against 48.6% for an observation-only control and 37.9% for an action-shuffled control. These simulation results support action-conditioned verification as a way to restore feedback during chunked execution while keeping the repair consistent with inference latency."
    },
    "groups": [
      "jul"
    ],
    "event_at": "2026-07-29T19:31:33+08:00"
  },
  {
    "id": "2510.06710v3",
    "title": "RLinf-VLA: A Unified and Efficient Framework for Reinforcement Learning of Vision-Language-Action Models",
    "abstract": "Recent studies have demonstrated the potential of reinforcement learning (RL) to improve the task performance of vision-language-action (VLA) models through interaction. However, current efforts remain fragmented, lacking a unified platform for fair comparison across architectures and algorithms, as well as an efficient system design for scalable training. Therefore, we present RLinf-VLA, a unified and efficient framework for scalable RL training of VLA models. RLinf-VLA standardizes the integration of diverse VLA architectures, RL algorithms, and heterogeneous simulators through a unified interface, enabling extensibility and reproducibility. To improve efficiency, the framework adopts a flexible resource allocation architecture for rendering, inference, and training in RL pipelines. In particular, RLinf-VLA introduces a hybrid fine-grained pipeline allocation strategy that achieves a 1.61$\\times$-1.88$\\times$ training speedup on ManiSkill. Using this framework, RL-trained models achieve strong performance across embodied benchmarks, including 98.11% success on 130 LIBERO tasks, 97.66% success on 25 ManiSkill tasks, and 84.63% average success across 6 RoboTwin tasks. In addition, RLinf-VLA distills a set of effective practices for RL-based VLA training. We envision RLinf-VLA as a foundational framework for efficient, unified, and reproducible research in embodied intelligence.",
    "history": "Submission history From: Mingjie Wei [ view email ] [v1] Wed, 8 Oct 2025 07:05:13 UTC (4,236 KB) [v2] Sat, 7 Feb 2026 10:46:31 UTC (3,630 KB) [v3] Wed, 12 Aug 2026 08:36:04 UTC (5,242 KB)",
    "meta": {
      "citation_title": "RLinf-VLA: A Unified and Efficient Framework for Reinforcement Learning of Vision-Language-Action Models",
      "citation_author": "Wang, Yu",
      "citation_date": "2025/10/08",
      "citation_online_date": "2026/08/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2510.06710",
      "citation_arxiv_id": "2510.06710",
      "citation_abstract": "Recent studies have demonstrated the potential of reinforcement learning (RL) to improve the task performance of vision-language-action (VLA) models through interaction. However, current efforts remain fragmented, lacking a unified platform for fair comparison across architectures and algorithms, as well as an efficient system design for scalable training. Therefore, we present RLinf-VLA, a unified and efficient framework for scalable RL training of VLA models. RLinf-VLA standardizes the integration of diverse VLA architectures, RL algorithms, and heterogeneous simulators through a unified interface, enabling extensibility and reproducibility. To improve efficiency, the framework adopts a flexible resource allocation architecture for rendering, inference, and training in RL pipelines. In particular, RLinf-VLA introduces a hybrid fine-grained pipeline allocation strategy that achieves a 1.61$\\times$-1.88$\\times$ training speedup on ManiSkill. Using this framework, RL-trained models achieve strong performance across embodied benchmarks, including 98.11% success on 130 LIBERO tasks, 97.66% success on 25 ManiSkill tasks, and 84.63% average success across 6 RoboTwin tasks. In addition, RLinf-VLA distills a set of effective practices for RL-based VLA training. We envision RLinf-VLA as a foundational framework for efficient, unified, and reproducible research in embodied intelligence."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-12T16:36:04+08:00"
  },
  {
    "id": "2511.07820v4",
    "title": "SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control",
    "abstract": "Despite the rise of billion-parameter foundation models trained across thousands of graphical processing units (GPUs), similar scaling gains have not been shown for humanoid control. Current neural controllers for humanoids remain modest in size, target a limited set of behaviors, and are trained on a handful of GPUs. We show that scaling model capacity, data, and compute yields a generalist humanoid controller capable of natural, robust whole-body movements. We position motion tracking as a scalable task for humanoid control, leveraging dense supervision from diverse motion-capture data to acquire human motion priors without manual reward engineering. We build a foundation model for motion tracking by scaling along three axes: network size (1.2M to 42M parameters), dataset volume (100M+ frames from 700 hours of motion capture), and compute (21k GPU hours). Beyond demonstrating the benefits of scale, we further show downstream utility through a real-time kinematic planner that bridges motion tracking to tasks such as navigation, enabling natural and interactive control, as well as a unified token space that supports virtual reality (VR) teleoperation and vision-language-action (VLA) models with a single policy. Through this interface, we demonstrate autonomous VLA-driven whole-body loco-manipulation requiring coordinated hand and foot placement. Scaling motion tracking exhibits favorable properties: performance improves steadily with compute and data diversity, and learned policies generalize to unseen motions, establishing motion tracking at scale as a practical foundation for humanoid control.",
    "history": "Submission history From: Zhengyi Luo [ view email ] [v1] Tue, 11 Nov 2025 04:37:40 UTC (16,929 KB) [v2] Thu, 4 Dec 2025 19:35:21 UTC (16,945 KB) [v3] Thu, 21 May 2026 17:26:49 UTC (27,316 KB) [v4] Thu, 13 Aug 2026 07:13:21 UTC (19,561 KB)",
    "meta": {
      "citation_title": "SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control",
      "citation_author": "Zhu, Yuke",
      "citation_doi": "10.1126/scirobotics.aed4592",
      "citation_date": "2025/11/11",
      "citation_online_date": "2026/08/13",
      "citation_pdf_url": "https://arxiv.org/pdf/2511.07820",
      "citation_arxiv_id": "2511.07820",
      "citation_abstract": "Despite the rise of billion-parameter foundation models trained across thousands of graphical processing units (GPUs), similar scaling gains have not been shown for humanoid control. Current neural controllers for humanoids remain modest in size, target a limited set of behaviors, and are trained on a handful of GPUs. We show that scaling model capacity, data, and compute yields a generalist humanoid controller capable of natural, robust whole-body movements. We position motion tracking as a scalable task for humanoid control, leveraging dense supervision from diverse motion-capture data to acquire human motion priors without manual reward engineering. We build a foundation model for motion tracking by scaling along three axes: network size (1.2M to 42M parameters), dataset volume (100M+ frames from 700 hours of motion capture), and compute (21k GPU hours). Beyond demonstrating the benefits of scale, we further show downstream utility through a real-time kinematic planner that bridges motion tracking to tasks such as navigation, enabling natural and interactive control, as well as a unified token space that supports virtual reality (VR) teleoperation and vision-language-action (VLA) models with a single policy. Through this interface, we demonstrate autonomous VLA-driven whole-body loco-manipulation requiring coordinated hand and foot placement. Scaling motion tracking exhibits favorable properties: performance improves steadily with compute and data diversity, and learned policies generalize to unseen motions, establishing motion tracking at scale as a practical foundation for humanoid control."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-13T15:13:21+08:00"
  },
  {
    "id": "2608.11739v1",
    "title": "G0.5: One Autoregressive Stream for Robot Reasoning and Action",
    "abstract": "The prevailing recipe for Vision-Language-Action (VLA) models couples a pretrained VLM with a separately trained flow-matching action expert. This makes the VLM a context encoder rather than a decision-maker. We introduce G0.5, a pretrained autoregressive VLA in which a single transformer decoder emits reasoning and action tokens under a single objective. Three components make this tractable at foundation-model scale: a learnable cross-embodiment action tokenizer that maps heterogeneous robot actions into a shared vocabulary; a native chain-of-thought stream interleaving task decomposition, object grounding, and action hints with action tokens; and a visual memory module that injects multi-second history through the vision encoder. Because reasoning and action share a single set of weights, the pretrained VLM's capabilities carry over to physical behavior: the model follows instructions closely, and prompts directly steer action granularity, task horizon, and out-of-distribution scene handling without further training. Pretrained on a large collection of robot datasets together with VQA samples, G0.5 surpasses state-of-the-art models across 7 independent regimes: real-world fine-tuning on R1lite and R1pro robots (76.7\\% vs.\\ 53.3\\% for $\\pi_{0.5}$ and 24.4\\% for GR00T-N1.7), the 2025 BEHAVIOR Challenge on 50 long-horizon household mobile manipulation tasks using a generalist policy (31.4\\% vs.\\ 26.3\\% for $\\pi_{0.5}$ and 26.1\\% for the challenge winner), DROID post-training followed by zero-shot transfer to an unseen environment and objects (82.5\\%), a language-following Pick-and-Place benchmark, LIBERO (98.9\\%), RoboTwin 2.0 (93.3\\%), and SimplerEnv-Bridge (87.3\\%).",
    "history": "Submission history From: Liu Yicheng [ view email ] [v1] Wed, 12 Aug 2026 07:26:47 UTC (24,803 KB)",
    "meta": {
      "citation_title": "G0.5: One Autoregressive Stream for Robot Reasoning and Action",
      "citation_author": "Zhao, Hang",
      "citation_date": "2026/08/12",
      "citation_online_date": "2026/08/12",
      "citation_pdf_url": "https://arxiv.org/pdf/2608.11739",
      "citation_arxiv_id": "2608.11739",
      "citation_abstract": "The prevailing recipe for Vision-Language-Action (VLA) models couples a pretrained VLM with a separately trained flow-matching action expert. This makes the VLM a context encoder rather than a decision-maker. We introduce G0.5, a pretrained autoregressive VLA in which a single transformer decoder emits reasoning and action tokens under a single objective. Three components make this tractable at foundation-model scale: a learnable cross-embodiment action tokenizer that maps heterogeneous robot actions into a shared vocabulary; a native chain-of-thought stream interleaving task decomposition, object grounding, and action hints with action tokens; and a visual memory module that injects multi-second history through the vision encoder. Because reasoning and action share a single set of weights, the pretrained VLM's capabilities carry over to physical behavior: the model follows instructions closely, and prompts directly steer action granularity, task horizon, and out-of-distribution scene handling without further training. Pretrained on a large collection of robot datasets together with VQA samples, G0.5 surpasses state-of-the-art models across 7 independent regimes: real-world fine-tuning on R1lite and R1pro robots (76.7\\% vs.\\ 53.3\\% for $\\pi_{0.5}$ and 24.4\\% for GR00T-N1.7), the 2025 BEHAVIOR Challenge on 50 long-horizon household mobile manipulation tasks using a generalist policy (31.4\\% vs.\\ 26.3\\% for $\\pi_{0.5}$ and 26.1\\% for the challenge winner), DROID post-training followed by zero-shot transfer to an unseen environment and objects (82.5\\%), a language-following Pick-and-Place benchmark, LIBERO (98.9\\%), RoboTwin 2.0 (93.3\\%), and SimplerEnv-Bridge (87.3\\%)."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-12T15:26:47+08:00"
  },
  {
    "id": "2608.30883v1",
    "title": "SleepWalking: Privileged Representation Shaping for End-to-End Blind Locomotion in Legged Robots",
    "abstract": "Partially observable locomotion requires a policy to act when task-relevant properties of the robot--environment state are not fully specified by instantaneous observations. Existing approaches often address this challenge by explicitly estimating missing physical variables or processing extended observation histories through structured architectures. We take a different view: partial observability is fundamentally an information-retention problem. The decisive question is not how task-relevant information enters the network, but whether the policy's internal state retains it. Guided by this perspective, we propose SleepWalking for Robot Locomotion (SWAQ), a one-stage end-to-end framework that uses next-step privileged physical reconstruction to shape what a recurrent history representation retains during policy learning, while the deployed actor uses only a direct history-to-action pathway. Under aligned training settings, SWAQ achieves a 15.0\\% higher peak mean terrain level than DWAQ, the strongest non-exteroceptive baseline, while using 44.4\\% fewer inference MACs per control step. Layerwise probes further show that information associated with the reconstructed physical variables remains linearly decodable through the policy head up to the layer preceding the action output. Complementary theoretical analysis relates privileged-variable recoverability to the achievable-return gap between history-based and privileged-information policy classes. These results suggest that semantic objectives can structure learning without requiring a corresponding architectural decomposition of the deployed controller.",
    "history": "Submission history From: Tenghui Wang [ view email ] [v1] Mon, 31 Aug 2026 14:42:03 UTC (8,530 KB)",
    "meta": {
      "citation_title": "SleepWalking: Privileged Representation Shaping for End-to-End Blind Locomotion in Legged Robots",
      "citation_author": "He, Liang",
      "citation_date": "2026/08/31",
      "citation_online_date": "2026/08/31",
      "citation_pdf_url": "https://arxiv.org/pdf/2608.30883",
      "citation_arxiv_id": "2608.30883",
      "citation_abstract": "Partially observable locomotion requires a policy to act when task-relevant properties of the robot--environment state are not fully specified by instantaneous observations. Existing approaches often address this challenge by explicitly estimating missing physical variables or processing extended observation histories through structured architectures. We take a different view: partial observability is fundamentally an information-retention problem. The decisive question is not how task-relevant information enters the network, but whether the policy's internal state retains it. Guided by this perspective, we propose SleepWalking for Robot Locomotion (SWAQ), a one-stage end-to-end framework that uses next-step privileged physical reconstruction to shape what a recurrent history representation retains during policy learning, while the deployed actor uses only a direct history-to-action pathway. Under aligned training settings, SWAQ achieves a 15.0\\% higher peak mean terrain level than DWAQ, the strongest non-exteroceptive baseline, while using 44.4\\% fewer inference MACs per control step. Layerwise probes further show that information associated with the reconstructed physical variables remains linearly decodable through the policy head up to the layer preceding the action output. Complementary theoretical analysis relates privileged-variable recoverability to the achievable-return gap between history-based and privileged-information policy classes. These results suggest that semantic objectives can structure learning without requiring a corresponding architectural decomposition of the deployed controller."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-31T22:42:03+08:00"
  },
  {
    "id": "2608.30536v1",
    "title": "Behavior-Skill: A Fine-Grained Benchmark for Evaluating Vision-Language-Action Policies in Long-Horizon Tasks",
    "abstract": "Reliable execution of long-horizon mobile manipulation tasks remains challenging because overall task success depends on the successful completion of multiple constituent skills. Existing benchmarks, however, still rely primarily on full-task rollouts and aggregate task-level metrics, making intermediate failures difficult to observe and analyze. We present Behavior-Skill, a benchmark that reformulates the learning and evaluation of long-horizon tasks around executable constituent skills. It contains 235,492 skill instances from 10,000 demonstrations across 50 household tasks and 34 semantic skill categories. Each instance pairs a skill instruction with an aligned observation-action segment, and is further associated with a restorable intermediate state and a skill success condition to enable independent evaluation under valid preconditions. We further introduce trajectory-level and skill-level metrics to characterize policy capability beyond aggregate task success. Extensive experiments across representative VLA policies including pi0.5 and GR00T on the complete 50-task benchmark show that failures are highly non-uniform across skills, with contact-rich manipulation skills forming persistent bottlenecks. These results demonstrate that Behavior-Skill complements full-task evaluation by exposing intermediate capability profiles for analyzing and improving long-horizon VLA policies. Behavior-Skill is publicly available at this https URL .",
    "history": "Submission history From: Chunyun Ma [ view email ] [v1] Mon, 31 Aug 2026 10:03:48 UTC (5,729 KB)",
    "meta": {
      "citation_title": "Behavior-Skill: A Fine-Grained Benchmark for Evaluating Vision-Language-Action Policies in Long-Horizon Tasks",
      "citation_author": "Chen, Xieyuanli",
      "citation_date": "2026/08/31",
      "citation_online_date": "2026/08/31",
      "citation_pdf_url": "https://arxiv.org/pdf/2608.30536",
      "citation_arxiv_id": "2608.30536",
      "citation_abstract": "Reliable execution of long-horizon mobile manipulation tasks remains challenging because overall task success depends on the successful completion of multiple constituent skills. Existing benchmarks, however, still rely primarily on full-task rollouts and aggregate task-level metrics, making intermediate failures difficult to observe and analyze. We present Behavior-Skill, a benchmark that reformulates the learning and evaluation of long-horizon tasks around executable constituent skills. It contains 235,492 skill instances from 10,000 demonstrations across 50 household tasks and 34 semantic skill categories. Each instance pairs a skill instruction with an aligned observation-action segment, and is further associated with a restorable intermediate state and a skill success condition to enable independent evaluation under valid preconditions. We further introduce trajectory-level and skill-level metrics to characterize policy capability beyond aggregate task success. Extensive experiments across representative VLA policies including pi0.5 and GR00T on the complete 50-task benchmark show that failures are highly non-uniform across skills, with contact-rich manipulation skills forming persistent bottlenecks. These results demonstrate that Behavior-Skill complements full-task evaluation by exposing intermediate capability profiles for analyzing and improving long-horizon VLA policies. Behavior-Skill is publicly available at https://github.com/nubot-nudt/Behavior-Skill."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-31T18:03:48+08:00"
  },
  {
    "id": "2608.29749v1",
    "title": "DriftingVLA: Native One-Step Vision-Language-Action Generation via Per-Dimension Temporal Drifting",
    "abstract": "Conventional flow-based vision-language-action (VLA) models support expressive continuous action generation but rely on multi-step refinement to produce each action chunk, increasing latency in online robot control. To address this issue, we introduce DriftingVLA, a native one-step VLA that generates a complete action chunk with a single action-expert forward pass. Rather than learning a flow field that requires iterative integration at inference, DriftingVLA uses a distribution-drifting objective to learn a direct noise-to-action-chunk mapping for one-step deployment. Since robot action dimensions carry distinct control semantics and distributional characteristics, we further introduce Per-Dimension Temporal Drifting (PDTD). PDTD treats the complete temporal trajectory of each action dimension as a separate drifting unit, enabling finer-grained modeling and shaping of dimension-specific action distributions. This per-dimension decomposition applies only to the training objective; the shared VLA model still generates the complete action chunk jointly, thereby preserving cross-dimensional dependencies. DriftingVLA achieves 98.32% success on LIBERO, 81.09% on RoboTwin 2.0, and 77.67% across six real-world single- and dual-arm tasks, outperforming the evaluated multi-step flow policy and one-step VLA baselines. Native one-step deployment also delivers a 3.36-fold speedup in action-chunk generation, eliminating iterative refinement without sacrificing control performance.",
    "history": "Submission history From: Yuxuan Gao [ view email ] [v1] Sun, 30 Aug 2026 12:21:44 UTC (13,334 KB)",
    "meta": {
      "citation_title": "DriftingVLA: Native One-Step Vision-Language-Action Generation via Per-Dimension Temporal Drifting",
      "citation_author": "Zhang, Yanyong",
      "citation_date": "2026/08/30",
      "citation_online_date": "2026/08/30",
      "citation_pdf_url": "https://arxiv.org/pdf/2608.29749",
      "citation_arxiv_id": "2608.29749",
      "citation_abstract": "Conventional flow-based vision-language-action (VLA) models support expressive continuous action generation but rely on multi-step refinement to produce each action chunk, increasing latency in online robot control. To address this issue, we introduce DriftingVLA, a native one-step VLA that generates a complete action chunk with a single action-expert forward pass. Rather than learning a flow field that requires iterative integration at inference, DriftingVLA uses a distribution-drifting objective to learn a direct noise-to-action-chunk mapping for one-step deployment. Since robot action dimensions carry distinct control semantics and distributional characteristics, we further introduce Per-Dimension Temporal Drifting (PDTD). PDTD treats the complete temporal trajectory of each action dimension as a separate drifting unit, enabling finer-grained modeling and shaping of dimension-specific action distributions. This per-dimension decomposition applies only to the training objective; the shared VLA model still generates the complete action chunk jointly, thereby preserving cross-dimensional dependencies. DriftingVLA achieves 98.32% success on LIBERO, 81.09% on RoboTwin 2.0, and 77.67% across six real-world single- and dual-arm tasks, outperforming the evaluated multi-step flow policy and one-step VLA baselines. Native one-step deployment also delivers a 3.36-fold speedup in action-chunk generation, eliminating iterative refinement without sacrificing control performance."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-30T20:21:44+08:00"
  },
  {
    "id": "2608.26239v1",
    "title": "WALL-SS: Scaling Long-horizon World Models via Next-Scale Autoregression",
    "abstract": "Generative world models provide robots with predictive models of how the world evolves under interaction, with growing potential for simulation, planning, policy evaluation, and robot learning. Beyond clip-level future prediction, a unified generative formulation should relate actions to consequences, support flexible horizons and continuous interaction, and enable reward-driven optimization. We introduce WALL-SS, a world model that generates visual futures through Scale-wise autoregressive Scaling, enabling action-controllable and long-horizon robotic simulation. WALL-SS represents embodied trajectories as causal sequences of temporally interleaved observations and actions, making action-dependent state transitions explicit while naturally supporting variable-length generation, streaming extension through reusable causal states, and direct optimization through sequence probabilities. To make this formulation effective over long horizons, we generate each future observation in a coarse-to-fine manner and develop three complementary components within the same hierarchy. Action-conditioned next-scale prediction injects scale-aligned action representations to improve action-future coupling and model both successful and failed behaviors. Scale-compressed long-horizon memory retains recent interactions at fine resolution while compressing distant observations and actions, with scale-wise dream forcing enhancing robustness to self-generated context. Finally, on-policy alignment optimizes autoregressive visual dynamics with action-following and long-term consistency rewards while preserving the pretrained visual distribution. Experiments show that WALL-SS improves action following and trajectory accuracy, supports coherent minute-long streaming rollout under bounded memory, and consistently benefits from on-policy alignment in reducing action drift and long-horizon inconsistency.",
    "history": "Submission history From: Yu Sun [ view email ] [v1] Wed, 26 Aug 2026 17:57:12 UTC (5,985 KB)",
    "meta": {
      "citation_title": "WALL-SS: Scaling Long-horizon World Models via Next-Scale Autoregression",
      "citation_author": "Wang, Qian",
      "citation_date": "2026/08/26",
      "citation_online_date": "2026/08/26",
      "citation_pdf_url": "https://arxiv.org/pdf/2608.26239",
      "citation_arxiv_id": "2608.26239",
      "citation_abstract": "Generative world models provide robots with predictive models of how the world evolves under interaction, with growing potential for simulation, planning, policy evaluation, and robot learning. Beyond clip-level future prediction, a unified generative formulation should relate actions to consequences, support flexible horizons and continuous interaction, and enable reward-driven optimization. We introduce WALL-SS, a world model that generates visual futures through Scale-wise autoregressive Scaling, enabling action-controllable and long-horizon robotic simulation. WALL-SS represents embodied trajectories as causal sequences of temporally interleaved observations and actions, making action-dependent state transitions explicit while naturally supporting variable-length generation, streaming extension through reusable causal states, and direct optimization through sequence probabilities. To make this formulation effective over long horizons, we generate each future observation in a coarse-to-fine manner and develop three complementary components within the same hierarchy. Action-conditioned next-scale prediction injects scale-aligned action representations to improve action-future coupling and model both successful and failed behaviors. Scale-compressed long-horizon memory retains recent interactions at fine resolution while compressing distant observations and actions, with scale-wise dream forcing enhancing robustness to self-generated context. Finally, on-policy alignment optimizes autoregressive visual dynamics with action-following and long-term consistency rewards while preserving the pretrained visual distribution. Experiments show that WALL-SS improves action following and trajectory accuracy, supports coherent minute-long streaming rollout under bounded memory, and consistently benefits from on-policy alignment in reducing action drift and long-horizon inconsistency."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-27T01:57:12+08:00"
  },
  {
    "id": "2608.26645v1",
    "title": "FLARE: A Failure-Aware Framework for Autonomous Correction and Recovery in Visual-Language Robotic Manipulation",
    "abstract": "Vision-Language-Action Models~(VLAs) have demonstrated significant promise in generalizing to complex, long-horizon robotic manipulation tasks. However, their performance remains brittle, as they are typically trained on trajectory-monotonic, failure-free demonstrations. This reliance on ``perfect\" data leaves them unable to recover from common execution errors, such as a missed grasp, a dropped object, or an unexpected collision. In this paper, we propose FLARE, a novel framework that endows VLAs with robust error recovery capabilities through a ``Retry\" and ``Reset\" paradigm. First, we introduce a ``Retry\" mechanism by injecting perturbation and bridging segments that decouple robot pose from environment state into demonstrations, enabling the policy to autonomously handle execution deviations. Second, to address critical, state-breaking (OOD) failures, we introduce a ``Reset\" pipeline. We leverage an MLLM for offline failure analysis to automatically identify OOD states from execution videos. This analysis enables the efficient, targeted collection of a small library of object-centric ``Reset\" skills, which are trained to restore the environment to a task-valid state. Our full framework integrates these learned policies. At inference, an online MLLM monitor arbitrates between task execution and ``Reset\" skills. Experiments on challenging, contact-rich manipulation tasks show our approach significantly improves task success and robustness.",
    "history": "Submission history From: Ganlong Zhao [ view email ] [v1] Thu, 27 Aug 2026 05:58:06 UTC (1,151 KB)",
    "meta": {
      "citation_title": "FLARE: A Failure-Aware Framework for Autonomous Correction and Recovery in Visual-Language Robotic Manipulation",
      "citation_author": "Li, Guanbin",
      "citation_date": "2026/08/27",
      "citation_online_date": "2026/08/27",
      "citation_pdf_url": "https://arxiv.org/pdf/2608.26645",
      "citation_arxiv_id": "2608.26645",
      "citation_abstract": "Vision-Language-Action Models~(VLAs) have demonstrated significant promise in generalizing to complex, long-horizon robotic manipulation tasks. However, their performance remains brittle, as they are typically trained on trajectory-monotonic, failure-free demonstrations. This reliance on ``perfect\" data leaves them unable to recover from common execution errors, such as a missed grasp, a dropped object, or an unexpected collision. In this paper, we propose FLARE, a novel framework that endows VLAs with robust error recovery capabilities through a ``Retry\" and ``Reset\" paradigm. First, we introduce a ``Retry\" mechanism by injecting perturbation and bridging segments that decouple robot pose from environment state into demonstrations, enabling the policy to autonomously handle execution deviations. Second, to address critical, state-breaking (OOD) failures, we introduce a ``Reset\" pipeline. We leverage an MLLM for offline failure analysis to automatically identify OOD states from execution videos. This analysis enables the efficient, targeted collection of a small library of object-centric ``Reset\" skills, which are trained to restore the environment to a task-valid state. Our full framework integrates these learned policies. At inference, an online MLLM monitor arbitrates between task execution and ``Reset\" skills. Experiments on challenging, contact-rich manipulation tasks show our approach significantly improves task success and robustness."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-27T13:58:06+08:00"
  },
  {
    "id": "2608.18701v1",
    "title": "SoftVTBench: A Deformation-Aware Visuo-Tactile Dataset and Benchmark for Deformable-Object Manipulation",
    "abstract": "Physical interaction quality is central to deformable-object manipulation, yet most benchmarks evaluate task success alone. A policy may complete the task while allowing slip or causing excessive compression. A primary bottleneck is the absence of visuo-tactile datasets that pair policy-visible contact observations with independent physical ground truth over complete tasks. We introduce SoftVTBench, a visuo-tactile dataset for physical-interaction-aware deformable-object manipulation. It contains 4,000 expert demonstrations and more than 50 assets, including volumetric deformable objects and visually matched rigid twins. At 20 Hz, each episode synchronizes multi-view RGB, dual-finger tactile RGB and marker motion, proprioception, language, and binary and continuous gripper actions, alongside evaluator-only finite-element (FEM) states. Building upon this dataset, we establish a closed-loop benchmark that uses fixed object-specific calibration to define the Deformation-aware Success Rate (DSR), which counts a rollout as successful only when it completes the task and keeps peak normalized deformation within tolerance. Across Diffusion Policy, $\\pi_{0.5}$, and FastWAM, all 12 in-distribution configurations contain successful rollouts that violate the deformation tolerance, accounting for 0.7--24% of each configuration's successes. Under distribution shift, visuo-tactile variants achieve higher task success in all six policy--suite comparisons and higher DSR in five, whereas their in-distribution benefits are mixed. These results show that making touch available does not by itself ensure effective multimodal fusion. SoftVTBench therefore provides a common visuo-tactile resource for studying not only whether a policy succeeds, but how it physically interacts with deformable objects and when touch improves that interaction.",
    "history": "Submission history From: Mingxin Wang [ view email ] [v1] Wed, 19 Aug 2026 08:58:47 UTC (8,964 KB)",
    "meta": {
      "citation_title": "SoftVTBench: A Deformation-Aware Visuo-Tactile Dataset and Benchmark for Deformable-Object Manipulation",
      "citation_author": "Yu, Haibao",
      "citation_date": "2026/08/19",
      "citation_online_date": "2026/08/19",
      "citation_pdf_url": "https://arxiv.org/pdf/2608.18701",
      "citation_arxiv_id": "2608.18701",
      "citation_abstract": "Physical interaction quality is central to deformable-object manipulation, yet most benchmarks evaluate task success alone. A policy may complete the task while allowing slip or causing excessive compression. A primary bottleneck is the absence of visuo-tactile datasets that pair policy-visible contact observations with independent physical ground truth over complete tasks. We introduce SoftVTBench, a visuo-tactile dataset for physical-interaction-aware deformable-object manipulation. It contains 4,000 expert demonstrations and more than 50 assets, including volumetric deformable objects and visually matched rigid twins. At 20 Hz, each episode synchronizes multi-view RGB, dual-finger tactile RGB and marker motion, proprioception, language, and binary and continuous gripper actions, alongside evaluator-only finite-element (FEM) states. Building upon this dataset, we establish a closed-loop benchmark that uses fixed object-specific calibration to define the Deformation-aware Success Rate (DSR), which counts a rollout as successful only when it completes the task and keeps peak normalized deformation within tolerance. Across Diffusion Policy, $\\pi_{0.5}$, and FastWAM, all 12 in-distribution configurations contain successful rollouts that violate the deformation tolerance, accounting for 0.7--24% of each configuration's successes. Under distribution shift, visuo-tactile variants achieve higher task success in all six policy--suite comparisons and higher DSR in five, whereas their in-distribution benefits are mixed. These results show that making touch available does not by itself ensure effective multimodal fusion. SoftVTBench therefore provides a common visuo-tactile resource for studying not only whether a policy succeeds, but how it physically interacts with deformable objects and when touch improves that interaction."
    },
    "groups": [
      "aug"
    ],
    "event_at": "2026-08-19T16:58:47+08:00"
  },
  {
    "id": "2606.01955v2",
    "title": "WALL-WM: Carving World Action Modeling at the Event Joints",
    "abstract": "WALL-WM is a World Action Model that shifts video-action learning from chunk-centric optimization to event-grounded Vision-Language-Action pretraining, using semantically coherent action events as the atomic unit of learning. Existing WAMs commonly initialize from multimodal or video foundation models and then optimize fixed-length action chunks conditioned directly on the current observation and instruction. Although convenient, this chunk-centric formulation creates a fundamental granularity mismatch. Language describes semantic goals and events, vision evolves through continuous scene dynamics, and actions operate at control-level timescales; forcing all three into the same fixed-length prediction window turns VLA training into short-horizon correlation fitting. WALL-WM addresses this mismatch by organizing both supervision and data around semantic events. Specifically, it pairs event-grounded VLA pretraining with a data ecosystem built from event-level captions and cluster-balanced sampling, enabling scalable learning over diverse behaviors, scenes, and task structures. From the same event-pretrained backbone, WALL-WM supports two complementary inference modes. The event mode consumes next-event descriptions and enables variable-length execution chunks, while the unified mode uses a VLM with Staircase Decoding to condition conventional fixed-length chunk inference while preserving a gradient-continuous VLA path. Together with Muon-optimizer-based large-scale pretraining infrastructure, WALL-WM provides a practical scale-up recipe for general-purpose WAMs. Experiments show that WALL-WM generalizes broadly across language, scenes, and tasks, achieving state-of-the-art performance in large-scale real-world generalization evaluation.",
    "history": "Submission history From: Yu Sun [ view email ] [v1] Mon, 1 Jun 2026 09:14:51 UTC (40,072 KB) [v2] Sun, 6 Sep 2026 05:36:17 UTC (40,073 KB)",
    "meta": {
      "citation_title": "WALL-WM: Carving World Action Modeling at the Event Joints",
      "citation_author": "Wang, Qian",
      "citation_date": "2026/06/01",
      "citation_online_date": "2026/09/06",
      "citation_pdf_url": "https://arxiv.org/pdf/2606.01955",
      "citation_arxiv_id": "2606.01955",
      "citation_abstract": "WALL-WM is a World Action Model that shifts video-action learning from chunk-centric optimization to event-grounded Vision-Language-Action pretraining, using semantically coherent action events as the atomic unit of learning. Existing WAMs commonly initialize from multimodal or video foundation models and then optimize fixed-length action chunks conditioned directly on the current observation and instruction. Although convenient, this chunk-centric formulation creates a fundamental granularity mismatch. Language describes semantic goals and events, vision evolves through continuous scene dynamics, and actions operate at control-level timescales; forcing all three into the same fixed-length prediction window turns VLA training into short-horizon correlation fitting. WALL-WM addresses this mismatch by organizing both supervision and data around semantic events. Specifically, it pairs event-grounded VLA pretraining with a data ecosystem built from event-level captions and cluster-balanced sampling, enabling scalable learning over diverse behaviors, scenes, and task structures. From the same event-pretrained backbone, WALL-WM supports two complementary inference modes. The event mode consumes next-event descriptions and enables variable-length execution chunks, while the unified mode uses a VLM with Staircase Decoding to condition conventional fixed-length chunk inference while preserving a gradient-continuous VLA path. Together with Muon-optimizer-based large-scale pretraining infrastructure, WALL-WM provides a practical scale-up recipe for general-purpose WAMs. Experiments show that WALL-WM generalizes broadly across language, scenes, and tasks, achieving state-of-the-art performance in large-scale real-world generalization evaluation."
    },
    "groups": [
      "sep"
    ],
    "event_at": "2026-09-06T13:36:17+08:00"
  },
  {
    "id": "2609.23103v1",
    "title": "DiagGen: Agentic Generation of Deformable Assets with Sim-based Diagnostics for Robotic Simulation",
    "abstract": "While simulation-ready deformable assets are essential for in-silico robotic manipulation tasks, existing generation frameworks typically assess physical plausibility after generation, leaving an object's simulated response unused as feedback for repairing upstream errors. We present DiagGen, an agentic framework that turns a single in-the-wild image into a simulation-ready deformable asset through a generate--simulate--diagnose--refine loop. DiagGen constructs part-aware geometry and material parameters, then uses a VLM (vision-language model)-based agent to select semantically informative regions, probe them in a physics simulator, observe material responses, and route evidence-backed repair cues to the responsible generation stage. Experiments on 40 assets show that diagnostics provides useful repair cues and can moderately improve the quality of generated deformable assets. Finally, we show that unlike assets generated from visual foundation models which may not be simulatable, DiagGen-generated deformables can be directly dropped into a high-fidelity physical simulator for the planning and simulation of contact-rich pick-and-place tasks. The project's website is this https URL .",
    "history": "Submission history From: Guanxiong Chen [ view email ] [v1] Sat, 19 Sep 2026 16:16:42 UTC (4,760 KB)",
    "meta": {
      "citation_title": "DiagGen: Agentic Generation of Deformable Assets with Sim-based Diagnostics for Robotic Simulation",
      "citation_author": "Shi, Fan",
      "citation_date": "2026/09/19",
      "citation_online_date": "2026/09/19",
      "citation_pdf_url": "https://arxiv.org/pdf/2609.23103",
      "citation_arxiv_id": "2609.23103",
      "citation_abstract": "While simulation-ready deformable assets are essential for in-silico robotic manipulation tasks, existing generation frameworks typically assess physical plausibility after generation, leaving an object's simulated response unused as feedback for repairing upstream errors. We present DiagGen, an agentic framework that turns a single in-the-wild image into a simulation-ready deformable asset through a generate--simulate--diagnose--refine loop. DiagGen constructs part-aware geometry and material parameters, then uses a VLM (vision-language model)-based agent to select semantically informative regions, probe them in a physics simulator, observe material responses, and route evidence-backed repair cues to the responsible generation stage. Experiments on 40 assets show that diagnostics provides useful repair cues and can moderately improve the quality of generated deformable assets. Finally, we show that unlike assets generated from visual foundation models which may not be simulatable, DiagGen-generated deformables can be directly dropped into a high-fidelity physical simulator for the planning and simulation of contact-rich pick-and-place tasks. The project's website is https://diaggen.github.io/."
    },
    "groups": [
      "w2"
    ],
    "event_at": "2026-09-20T00:16:42+08:00"
  },
  {
    "id": "2610.04767v1",
    "title": "Flow Policies as Actions of Skill-Level World Models: Learned and Symbolic Abstractions for Long-Horizon Planning",
    "abstract": "Latent world models enable robots to plan by predicting the consequences of actions. Planning long tasks with control-rate actions requires many prediction steps, which enlarges the search space and accumulates error. Skill-level actions shorten these sequences, but a symbolic skill vocabulary requires domain knowledge and labeled demonstrations. We construct skill-level actions from the inputs of a flow-matching policy trained on demonstrations segmented into complete skills. The policy maps a noise seed and an observation, optionally with a code or label, to a complete skill execution, so one execution is one world-model transition. On this mechanism we propose four action abstractions with increasing task knowledge: a compressed seed, two discrete codes learned from the demonstrations, and a symbolic label. We evaluate them with a common world-model training procedure and planning framework on simulated block rearrangement tasks that require up to 14 sequential skills. The symbolic label succeeds in over 90% of the tasks that require up to six skills and degrades beyond. Without any label, an object-centric learned code matches it on single-skill tasks and retains half to three quarters of its success on tasks of two to five skills. Ablations attribute much of the label's advantage to its planner knowing which actions are applicable, rather than to the label itself. Beyond six skills the search, not the world model, limits success. Project page: this https URL",
    "history": "Submission history From: Andreu Matoses Gimenez [ view email ] [v1] Sat, 3 Oct 2026 21:07:46 UTC (3,867 KB)",
    "meta": {
      "citation_title": "Flow Policies as Actions of Skill-Level World Models: Learned and Symbolic Abstractions for Long-Horizon Planning",
      "citation_author": "Alonso-Mora, Javier",
      "citation_date": "2026/10/03",
      "citation_online_date": "2026/10/03",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.04767",
      "citation_arxiv_id": "2610.04767",
      "citation_abstract": "Latent world models enable robots to plan by predicting the consequences of actions. Planning long tasks with control-rate actions requires many prediction steps, which enlarges the search space and accumulates error. Skill-level actions shorten these sequences, but a symbolic skill vocabulary requires domain knowledge and labeled demonstrations. We construct skill-level actions from the inputs of a flow-matching policy trained on demonstrations segmented into complete skills. The policy maps a noise seed and an observation, optionally with a code or label, to a complete skill execution, so one execution is one world-model transition. On this mechanism we propose four action abstractions with increasing task knowledge: a compressed seed, two discrete codes learned from the demonstrations, and a symbolic label. We evaluate them with a common world-model training procedure and planning framework on simulated block rearrangement tasks that require up to 14 sequential skills. The symbolic label succeeds in over 90% of the tasks that require up to six skills and degrades beyond. Without any label, an object-centric learned code matches it on single-skill tasks and retains half to three quarters of its success on tasks of two to five skills. Ablations attribute much of the label's advantage to its planner knowing which actions are applicable, rather than to the label itself. Beyond six skills the search, not the world model, limits success. Project page: https://andreumatoses.github.io/research/flow-skill-wm"
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-10-04T05:07:46+08:00"
  },
  {
    "id": "2610.04607v1",
    "title": "ForeAct3D: Policy-Grounded Future World Modeling for VLA Policies",
    "abstract": "Robots need to anticipate how their actions will change the world, since manipulation success hinges on the resulting contacts and object motions. However, existing Vision-Language-Action (VLA) policies that predict future observations from shared features leave the forecast decoupled from the actions the policy will actually execute, and impose no physical constraints on how the scene may evolve. We introduce ForeAct3D, a framework for policy-grounded future world modeling within VLA policies. Learnable geometric queries decode depth, semantic segmentation, and camera pose from the policy representation into current and future semantic 3D scene states, and the future queries are conditioned on the policy-generated action chunk to ground the forecast in the planned interaction. A physical-consistency closure relates the two states through background staticity and instance-level rigidity, and anchors the wrist-camera pose to end-effector kinematics. These objectives shape the shared representation used for action generation during training, and no future prediction is required at inference. Without robot pretraining, ForeAct3D achieves 98.3\\% average success on LIBERO and an average task length of 3.73 on CALVIN, outperforming its base policy on every suite. Ablations show that semantic 3D supervision, physical consistency, and action conditioning each improve manipulation performance, and that action conditioning substantially improves future object localization. Real-world experiments on spatial placement, object insertion, and sequential manipulation further raise average success from 6.7\\% to 37.8\\% over the base policy. The project page and code are available at this https URL .",
    "history": "Submission history From: Zhe Tao [ view email ] [v1] Sat, 3 Oct 2026 15:49:04 UTC (3,538 KB)",
    "meta": {
      "citation_title": "ForeAct3D: Policy-Grounded Future World Modeling for VLA Policies",
      "citation_author": "Yan, Yan",
      "citation_date": "2026/10/03",
      "citation_online_date": "2026/10/03",
      "citation_pdf_url": "https://arxiv.org/pdf/2610.04607",
      "citation_arxiv_id": "2610.04607",
      "citation_abstract": "Robots need to anticipate how their actions will change the world, since manipulation success hinges on the resulting contacts and object motions. However, existing Vision-Language-Action (VLA) policies that predict future observations from shared features leave the forecast decoupled from the actions the policy will actually execute, and impose no physical constraints on how the scene may evolve. We introduce ForeAct3D, a framework for policy-grounded future world modeling within VLA policies. Learnable geometric queries decode depth, semantic segmentation, and camera pose from the policy representation into current and future semantic 3D scene states, and the future queries are conditioned on the policy-generated action chunk to ground the forecast in the planned interaction. A physical-consistency closure relates the two states through background staticity and instance-level rigidity, and anchors the wrist-camera pose to end-effector kinematics. These objectives shape the shared representation used for action generation during training, and no future prediction is required at inference. Without robot pretraining, ForeAct3D achieves 98.3\\% average success on LIBERO and an average task length of 3.73 on CALVIN, outperforming its base policy on every suite. Ablations show that semantic 3D supervision, physical consistency, and action conditioning each improve manipulation performance, and that action conditioning substantially improves future object localization. Real-world experiments on spatial placement, object insertion, and sequential manipulation further raise average success from 6.7\\% to 37.8\\% over the base policy. The project page and code are available at https://github.com/anthonytao80-crypto/ForeAct3D."
    },
    "groups": [
      "w4"
    ],
    "event_at": "2026-10-03T23:49:04+08:00"
  }
]