Skip to content
#Robot models and learning
Oct 22Thu
  1. arXiv cs.RO57

    SP-HLM: Sparse-Pose-Driven Motion Generation and Whole-Body Control Framework for Humanoid Loco-Manipulation

    SP-HLM proposes a sparse-pose-driven interface framework that uses five task-critical body poses plus a learned implicit motion prior to represent spatial intent and whole-body coordination in complementary forms. The SP-VLA vision-language-action model first predicts task-relevant sparse poses and finger joint positions from multimodal observations and language instructions via asynchronous action-chunk inference. The SP-MGM motion generation model, trained through two-stage distillation, reconstructs coordinated full-body motion in real time, and the SP-GMT general motion tracker then combines the explicit sparse targets with the implicit motion prior for whole-body control.

  4. arXiv cs.RO51

    Paper proposes path to improve robot zero-shot generalization through data diversity

    An arXiv paper formalizes zero-shot generalization in physical AI as coverage of the deployment distribution, noting that more samples from the same collector can only slowly shrink the unseen region, while each additional contributor can expand coverage. Based on this, the paper proposes a shared latent space for robot actions that combines multimodal alignment, generative augmentation, and noisy-label learning, and suggests that developers build applications around the standard robot Origmi to provide diverse data. It concludes with a testable hypothesis and evaluation protocol linking real-robot zero-shot success rates to data diversity.

  5. arXiv cs.RO67

    SpatialDex: Pre-training 3D-Grounded Bimanual Dexterous Manipulation from RGB-Only Human Videos

    SpatialDex is a framework that learns 3D-grounded bimanual dexterous manipulation from RGB-only human videos through reconstruction-augmented visuomotor pre-training, learning human behavior together with its spatial context. It reconstructs scene geometry and estimates hand motion from each video, aligns the two at a video-level scale using the hand-size prior of 3D hand estimates, and uses the reconstructed geometry as both a policy input and a prediction target.

  6. arXiv cs.RO57

    CausGround: Supporting Adaptive Embodied Planning with Causal Interaction Knowledge

    CausGround proposes a non-parametric causal interaction grounding framework that distills interaction experience into reusable, action-centered causal knowledge, identifies action preconditions and effects, and organizes them into a factorized world-state representation. At planning time, it aligns task-relevant causal dependencies with current observations, supporting backward reasoning over preconditions and forward reasoning over effects.

  7. arXiv cs.RO55

    FlowVLA: Flow-Aware Physical Completion Estimation for Quantitative Robotic Pouring

    FlowVLA separates action generation from physical-completion estimation. A shared π0.5 action expert learns approach, pour, and return, while a target-independent estimator fuses visual, joint, wrist-wrench, and recurrent force-belief histories to estimate transferred mass. It triggers return when this estimate plus predicted residual outflow reaches the requested mass.

  10. arXiv cs.RO60

    End-to-End Reinforcement Learning for Quadrupedal Jumping in Lunar Gravity

    The study proposes a multistage reinforcement learning method that produces a single end-to-end control policy for commanded point-to-point jumping by a quadruped robot under lunar gravity. The policy architecture is simple and relies only on mission-compatible onboard sensing; in simulation, landing errors are smaller than the robot's body size. The work also analyzes torque and power demands and characterizes performance degradation under extreme outlier events, including control-update freezes of up to 1 second during flight. In hardware transfer tests, the policy successfully completed a vertical jump under lunar-gravity offloading, and measured torque and power profiles matched the simulation and stayed within hardware limits.

  12. arXiv cs.RO55

    Distance-Field-Based Reinforcement Learning for Reactive and Robust Obstacle Avoidance in Robotic Arms

    This paper proposes a reinforcement learning framework that integrates the Euclidean Distance Field (EDF) directly into the observation space, providing the agent with continuous geometric cues to achieve stable and responsive obstacle avoidance behavior and simplifying reward design; it effectively avoids obstacles both with and without artificial potential field guidance.

  13. arXiv cs.RO56

    PAL: A Hierarchical Memory Framework with Plan-First, Act-Later Streaming Reasoning for Robot Manipulation

    The paper proposes PAL (Plan first, Act Later), a hierarchical memory framework for vision-language robot manipulation. A high-level vision-language model (VLM) converts streaming visual observations into a subtask_list, and at execution time selects the currently relevant subtask from it to pass to a low-level vision-language-action (VLA) policy.

  15. arXiv cs.RO49

    Relational Geometry Forcing: Improving Robot Manipulation Learning in World Action Models (WAMs) via Relational Geometry Constraints

    The paper proposes Relational Geometry Forcing (RGF), which injects geometric priors from a frozen 3D foundation model into World Action Models (WAMs) by aligning pairwise feature relations instead of pointwise feature alignment, avoiding the degradation of action learning caused by dense geometric targets over-constraining compact action features.

  16. arXiv cs.RO65

    TIDE: An online post-training system for human-in-the-loop deployment in robot manipulation

    The authors propose TIDE (Training In DEployment), a human-in-the-loop online post-training system that uses a provenance-conditioned supervised learning framework to update the policy online from human interventions and policy rollouts during deployment, without learning a critic, reward model, or action scorer.

  17. arXiv cs.RO60

    DepState-JEPA: Depth-Grounded State Representation Learning for Visual Perception and Manipulation

    The authors propose DepState-JEPA, a dual-stream RGB-D joint-embedding predictive architecture that connects spatial understanding with action-conditioned prediction through a shared latent state, DepState. Built on V-JEPA 2.1, the framework uses depth both as a complementary observation stream and as the grounding for metric 3D space, and constrains state learning with ray-conditioned readouts, object-trajectory supervision, and calibrated forward kinematics.

  18. arXiv cs.RO64

    SMR: Learning streaming motion retargeting for humanoids from human semantics

    SMR is a streaming neural motion retargeter learned from semantic-conditioned physics experts. It converts human motion into robot task-space targets, which the physics experts track in simulation, and then uses a strictly causal Transformer to learn their executed trajectories. Each new frame uses only current and past human observations to generate a robot motion reference for a downstream whole-body controller. Experiments show that the semantic-conditioned physics experts outperform direct-optimization retargeting in semantic fidelity and physical validity, and that the resulting streaming retargeter performs better in closed loop under a unified whole-body controller than kinematic and physics-refinement baselines.

  19. arXiv cs.RO55

    How Representational Priors Shape Generalization in World Action Models

    This study proposes that the generalization performance of world action models (WAMs) diverges significantly depending on the representational geometry of their generative backbones, and quantifies this prior using nuisance orthogonality (the degree to which features are orthogonal to the in-distribution manifold under out-of-distribution visual variations). Experiments across various image and video backbones find that higher orthogonality before training is associated with stronger robustness to visual variations after WAM training, enabling prospective prediction of policy generalization before training. Analysis of robust WAMs also reveals suppressed feature displacement and robust action generation.

  21. arXiv cs.RO50

    State-dependent objective coordination enables end-to-end balance and navigation for two-wheeled robots

    An arXiv paper proposes a lightweight state-dependent objective coordination method for end-to-end balance and navigation of two-wheeled inverted-pendulum robots. The method introduces a pitch-angle-conditioned scalar coefficient β(s)∈[0,1] that dynamically adjusts the relative weight of balance stabilization and navigation objectives during training, without requiring an additional control hierarchy or attitude-stabilization controllers such as PID or LQR.

  23. arXiv cs.RO55

    SLoSa: A Small, Low-Power, Safety-Aware Foundation Model Framework for Robotic Task Planning

    SLoSa is a neuro-symbolic compression and fine-tuning framework that builds a safety-aware dataset using classical planners, a teacher large language model, and formal verifiers, then applies supervised fine-tuning and reinforcement learning fine-tuning guided by a verification-based reward machine to train small, low-power, safety-aware robotic task planning models.

  24. arXiv cs.RO47

    One Policy, Many Styles: A Comparative Study of Preference Aggregation Locations in Steerable Humanoid Control

    A study compared four preference aggregation methods—reward-level, advantage-level, loss-level, and PCGrad—across two humanoid robot control frameworks, finding that no aggregation level is optimal across all environments and metrics. The experiments used Preference Reward Responsiveness (PRR) combined with coverage and utility metrics; results show that early aggregation produces a wider range of behavioral responses but introduces more unintended objective interference, while later aggregation improves control precision at the cost of some responsiveness.

  27. arXiv cs.RO51

    inr2action: Encoding Robot Episodes as Neural Implicit Representations for Recall and Action

    The paper proposes the inr2action framework, which fits a neural implicit representation to each robot manipulation episode starting from a shared meta-learned initialization; the resulting weight delta serves as memory, with a size that does not vary with episode length and stays fixed from one hundred frames to one hundred thousand frames.

  28. arXiv cs.RO72

    Real2Sem2Real: Generating Cross-Instance Robot Training Data from a Single Human RGB Demonstration

    Real2Sem2Real (R2Sem2R) proposes a real-semantic-real pipeline that turns a single monocular RGB human demonstration video into thousands of robot demonstrations covering multiple semantic object instances within the same category. The method tracks object SE(3) poses using DINOv3-distilled Gaussian splats, outperforming existing methods and FoundationPose with estimated metric depth, and maps the human hand to a feasible robot grasp that can be transferred to digital twins.

  30. arXiv cs.RO44

    Meta-Learning-Based Mode Recognition and Adaptive Control for Human-Robot Interaction

    The framework uses a meta-learning-trained neural network to recognize two human-robot interaction modes, active guidance and passive following, without manually setting switching thresholds. The passive mode recognition rate rose from 90.1% to 99.1%, and the active mode from 79.8% to 97.1%. An adaptive meta-learning controller can compensate for human disturbances and adapt online to varying interaction forces. A multi-trajectory meta-learning algorithm jointly learns an environment basis representation and force-related parameters to handle unmodeled dynamics. Experiments verified its robust control and reliable mode recognition performance.

  31. HuggingFace Daily Papers71

    DONIC: Generalist Dexterous Bimanual Tracking Control Learned from Human Demonstrations

    The paper presents DONIC, a generalist bimanual dexterous tracker jointly trained across manipulation categories and human motion sources, and releases BiHOI, a multi-view dataset containing 1,575 sequences covering 229 objects, emphasizing sustained bimanual interaction and retargeting reconstructed interactions into robot references.

  32. HuggingFace Daily Papers47

    Memory-Centric Continual Learning (MCCL): Memory-Driven Continual Learning for AI Agents

    The paper proposes Memory-Centric Continual Learning (MCCL), arguing that memory should be judged by whether it improves an agent's subsequent actions rather than by fact retrieval. It includes a three-axis taxonomy (organizing memory by storage location, form, and management approach), a two-loop lifecycle, an evaluation protocol, and a memory scaling framework.

  33. HuggingFace Daily Papers62

    AW-0: An Agentic World Action Model Framework Coupling Code Planning with Video Action Prediction

    This work proposes AW-0, an agentic world action model framework that models code, future video, and robot actions as three coupled generative streams within a unified mixture-of-transformers architecture. A code expert generates parameterized skill routines and control logic, while video and action experts generate future visual states and low-level motor commands.

  35. 机器人全球资讯 · 微信公众号58

    Embodied AI company Gamma Intelligent, incubated by China Telecom, established in Shanghai

    The China Telecom Artificial Intelligence Research Institute (TeleAI) officially announced on October 20 that Gamma Robtics, its first incubated market-oriented innovation company, was recently registered and established in Xuhui District, Shanghai. The company focuses on full-model innovation and applications for self-evolving humanoid robots and will jointly build a humanoid robot laboratory with TeleAI.

  37. HuggingFace Daily Papers62

    Echo: Jointly Denoising Action and Task Status Signals for Vision-Language-Action Models

    Echo proposes a framework that treats two task status signals, stop and failure, as two dimensions of each action chunk and denoises them together with robot actions, thereby distinguishing ongoing execution, successful completion, and failure. The method requires no frame-level progress annotations; it provides complementary supervision by constructing successful, failed, and recovery trajectories, and it lets status signals denoise faster than actions so that subsequent action denoising steps can build on cleaner status estimates. Real-world manipulation experiments show that Echo accurately predicts task status and consistently improves policy performance.

  38. HuggingFace Daily Papers67

    FocusVLA: A VLA Model with Future-Intent-Guided Dynamic View Routing

    FocusVLA proposes a stage-aware multi-view routing framework that predicts future interaction intent from historical actions and uses it to dynamically coordinate global and local views during action generation. On RoboTwin 2.0, its average success rate improves by 10.10 percentage points over the strongest baseline, π_{0.5}; in real-world evaluation, the success rate rises from 71.7% to 85.0%. The framework's intent prediction and dynamic routing modules can be integrated into existing VLA frameworks without modifying the core architecture.

  39. HuggingFace Daily Papers58

    HiSTEVE 1.0: A Hierarchical VLA Agent for Open-World Games

    HiSTEVE 1.0 is a hierarchical VLA agent that achieves long-horizon autonomy and native keyboard-and-mouse control in Minecraft through a capability-aware agent harness. Its orchestration layer uses structured memory to track progress, with a VLM planner selecting subtasks within the VLA's capabilities; the control layer uses a single VLA to execute all subtasks with vision-only perception.