Skip to content

待精读队列

机制：每日抓取自动追加新论文；Agent 逐篇精读，产出【核心思想/方法拆解/为什么重要/一句话】，读完把 [ ] 勾成 [x]。原则：Insight，不流水账。

待精读

  • [x] Revisiting the "Push-T" Robot Manipulation Task with Agentic Robotics · https://arxiv.org/abs/2608.18227 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] Zero-Shot Transfer of Force Map Estimation Across GelSight Mini Sensors · https://arxiv.org/abs/2608.18240 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] LabDex: A Hierarchical Benchmark for Dexterous Manipulation in Laboratories · https://arxiv.org/abs/2608.18618
  • [x] Orienteering Problem with Uncertain Time-Varying Rewards: Framework and Benchmark for Everyday Service Robotics · https://arxiv.org/abs/2608.18672 〔归位·越界跳过〕
  • [x] SoftVTBench: A Deformation-Aware Visuo-Tactile Dataset and Benchmark for Deformable-Object Manipulation · https://arxiv.org/abs/2608.18701
  • [x] The Missing Touch: Spatially Distributed Tactile Feedback Brings Teleoperation Closer to Human Dexterity · https://arxiv.org/abs/2608.19372
  • [x] SCAPE: Scenario-Conditioned Simulation-Augmented Policy Evaluation · https://arxiv.org/abs/2608.19425 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] CoToGrasp: Contact-Topology-Conditioned Dexterous Grasp Synthesis via Canonical Workspace Learning · https://arxiv.org/abs/2608.19776 〔归位·越界跳过〕
  • [x] Wave-Based Bilateral Teleoperation between Nonlinear Manipulators with Direct Contact Force Feedback · https://arxiv.org/abs/2608.20043 〔归位·越界跳过〕
  • [x] DECOWAM: Decoupled Whole-Body World-Action Model for Legged Mobile Manipulation · https://arxiv.org/abs/2608.20114 〔归位·越界跳过〕
  • [x] Video2DoorTraversal: Push Door Traversal via Simulated Door Twins · https://arxiv.org/abs/2608.20251 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] Koala Gripper: Co-designing Robotic Grippers and Data-Capture Devices for Scaling Dexterous Manipulation Learning · https://arxiv.org/abs/2608.20546
  • [x] Scalable Distributed Simulation-Based Testing for Automated Driving Systems · https://arxiv.org/abs/2608.20904 〔归位·越界跳过〕
  • [x] ViTacPhys: Physical Property-Aware Grasping from Human Visual-Tactile Demonstrations · https://arxiv.org/abs/2608.21355
  • [x] In-Situ Reconstruction of the International Space Station Using 3D Gaussian Splatting and Astrobee · https://arxiv.org/abs/2608.21685 〔归位·越界跳过〕
  • [x] Safety-Critical Bilateral Teleoperation for Omnidirectional Aerial Manipulation Using Force-Sensorless Haptic Feedback · https://arxiv.org/abs/2608.21735 〔归位·越界跳过〕
  • [x] GuardianBench: A Same-Scene Instruction-Contrastive Benchmark for Latent Contextual Risk in Embodied AI · https://arxiv.org/abs/2608.21928 〔归位·越界跳过〕
  • [x] Contact-Rich Robotic Manipulation in Construction via Zero-Shot Learning: A Diffusion Policy-Guided Adaptive Control · https://arxiv.org/abs/2608.22100 〔归位·越界跳过〕
  • [x] BehaviorWorldGen: Closing the Loop between Action Models and World Simulators via Controllable Behavior-Aware Structured World Generation · https://arxiv.org/abs/2608.22187
  • [x] The Imitator Game: Benchmarking Robot Imitative Ability Beyond Action Prediction · https://arxiv.org/abs/2608.22301
  • [x] Macro-Operator Generation and Predicate Selection for TAMP Operator Learning · https://arxiv.org/abs/2608.23629 〔归位·越界跳过〕
  • [x] Enhancing Sim2Real Transfer for Torque-Controlled Robots through Real2Sim Dynamics Estimation and Reinforcement Learning · https://arxiv.org/abs/2608.22629 〔归位·越界跳过〕
  • [x] A Statistical Audit of Physical AI Benchmark Redundancy · https://arxiv.org/abs/2608.25940
  • [x] Point2Pose: Occlusion-Recovering 6D Pose Tracking and 3D Reconstruction for Multiple Unknown Objects Via 2D Point Trackers · https://arxiv.org/abs/2604.10415 〔归位·越界跳过〕
  • [x] Regularized Latent Dynamics Prediction is a Strong Baseline For Behavioral Foundation Models · https://arxiv.org/abs/2603.15857 〔归位·越界跳过〕
  • [x] ScaRF-SLAM: Scale-Consistent Reconstruction with Feed-Forward Models and Classical Visual SLAM · https://arxiv.org/abs/2606.00307 〔归位·越界跳过〕
  • [x] UNRIO: Uncertainty-Aware Velocity Learning for Radar-Inertial Odometry · https://arxiv.org/abs/2604.13584 〔归位·越界跳过〕
  • [x] Scene2Demo: Self-Evolving Embodied Data Generation via Object-Action Graph · https://arxiv.org/abs/2602.12065
  • [x] Human vs. Teleoperated Robots in Vineyard Management: A Simulation-Based Analysis of Travel Speed, Routing, and Task Performance · https://arxiv.org/abs/2507.04167 〔归位·越界跳过〕
  • [x] Choose Your Game Wisely: Measuring Game-Theoretic Structures in Real-World Vehicle Interactions · https://arxiv.org/abs/2608.25917 〔归位·越界跳过〕
  • [x] Trust-Aware Sequential Decision Making and Rollout Planning for Resilient Multi-Robot Systems · https://arxiv.org/abs/2608.25690 〔归位·越界跳过〕
  • [x] Saliency-Depth Conditioning for Zero-Shot Segmentation of Communication-Tower Components in Cluttered UAV Imagery · https://arxiv.org/abs/2608.25435 〔归位·越界跳过〕
  • [x] Longitudinal Robot Learning from Demonstration with Care Providers in a Home Environment · https://arxiv.org/abs/2608.25196 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] CRESSim-Neo: A Batched GPU Simulation Engine for Surgical Robotics and Robot Learning · https://arxiv.org/abs/2608.25192 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] SkyDrive: Learning to Drive in a New City from Aerial Traffic Monitoring · https://arxiv.org/abs/2608.25142 〔归位·越界跳过〕
  • [x] Extending Ground-Constraint LiDAR-IMU Calibration to Tilted Surfaces in a Continuous-Time Framework · https://arxiv.org/abs/2608.25135 〔归位·越界跳过〕
  • [x] ROS2 Connect: A new ROS2 over WAN Solution · https://arxiv.org/abs/2608.25102 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · HABIT: Human-Aware Behavior and Interaction Training Dataset for Robot Manipulation · https://arxiv.org/abs/2606.31682 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · Open-AoE: An Open Egocentric Manipulation Dataset and Toolchain for Embodied Learning · https://arxiv.org/abs/2607.14183
  • [x] 2026 · SurgSync: Time-Synchronized Multi-Modal Data Collection Framework and Dataset for Surgical Robotics · https://arxiv.org/abs/2603.06919 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · CRAFT: Video Diffusion for Bimanual Robot Data Generation · https://arxiv.org/abs/2604.03552 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · Dexora: Open-source VLA for High-DoF Bimanual Dexterity · https://arxiv.org/abs/2605.18722 〔归位·越界跳过〕
  • [x] 2026 · Worlds in One Demo: A Synthetic Data Engine for Learning Open-World Mobile Manipulation · https://arxiv.org/abs/2607.13154 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · Learning While Deploying: Fleet-Scale Reinforcement Learning for Generalist Robot Policies · https://arxiv.org/abs/2605.00416 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · COBALT: Crowdsourcing Robot Learning via Cloud-Based Teleoperation with Smartphones · https://arxiv.org/abs/2605.19138
  • [x] 2026 · ACME: A Multi-Cultural, Multi-Embodiment Social-Navigation Dataset · https://arxiv.org/abs/2607.21964 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · SIEVE: Structure-Aware Data Selection for Imitation Learning with VLA Models · https://arxiv.org/abs/2607.06442
  • [x] 2026 · DexVerse: A Modular Benchmark for Multi-Task, Multi-Embodiment Dexterous Manipulation · https://arxiv.org/abs/2607.08751 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · EaDex: A Cross-Embodiment Dexterous Manipulation Framework from Low-Cost Demonstrations · https://arxiv.org/abs/2606.03268 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · UniDexTok: A Unified Dexterous Hand Tokenizer from Real Data · https://arxiv.org/abs/2606.10683 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · HRDexDB: A Paired Human-Robot Dataset for Cross-Embodiment Dexterous Grasping · https://arxiv.org/abs/2604.14944
  • [x] 2026 · EgoInfinity: A Web-Scale 4D Hand-Object Interaction Data Engine for Any-View Robot Retargeting and Video-to-Action Robot Learning · https://arxiv.org/abs/2606.17385
  • [x] 2026 · AXIS: A Growable Community-Driven Data Engine for Scalable Robot Manipulation · https://arxiv.org/abs/2607.21588
  • [x] 2026 · \(μ_0\): A Scalable 3D Interaction-Trace World Model · https://arxiv.org/abs/2606.13769 〔归位·越界跳过〕
  • [x] 2026 · YUBI: Yielding Universal Bidigital Interface for Bimanual Dexterous Manipulation at Scale · https://arxiv.org/abs/2606.10244
  • [x] 2026 · Diversity You Can Actually Measure: A Fast, Model-Free Diversity Metric for Robotics Datasets · https://arxiv.org/abs/2603.11634
  • [x] 2026 · \(π_{0.7}\): a Steerable Generalist Robotic Foundation Model with Emergent Capabilities · https://arxiv.org/abs/2604.15483 〔归位·越界跳过〕
  • [x] 2026 · Vision-Language-Action in Robotics: A Survey of Datasets, Benchmarks, and Data Engines · https://arxiv.org/abs/2604.23001 〔精读·见 深度精读-数据方向剩余批次〕
  • [x] 2026 · Turning Video Models into Generalist Robot Policies · https://arxiv.org/abs/2605.27817
  • [x] 2026 · KITE: Decoupling Kinematics and Interaction for Zero-Shot Cross-Embodiment Manipulation · https://arxiv.org/abs/2606.22113
  • [x] 2026 · From Passive Video to Editable Experience: Physically Grounded Experience Synthesis for Embodied Intelligence · https://arxiv.org/abs/2607.26903
  • [x] 2026 · Data Analogies Enable Efficient Cross-Embodiment Transfer · https://arxiv.org/abs/2603.06450
  • [x] 2026 · Robots Need More than VLA and World Models · https://arxiv.org/abs/2606.06556
  • [x] 2026 · Co-training with Ego-centric Video and Demonstration for Robot Navigation Task · https://arxiv.org/abs/2606.01951
  • [x] 2026 · Handroid: Bridging Dexterous Hand and Humanoid · https://arxiv.org/abs/2607.16187
  • [x] 2026 · Cloak: Zero-Shot Cross-Embodiment Manipulation by Masking the End-Effector from the VLA · https://arxiv.org/abs/2606.22836
  • [x] 2026 · Motion-Focused Latent Action Enables Cross-Embodiment VLA Training from Human EgoVideos · https://arxiv.org/abs/2606.18955
  • [x] 2026 · Cross-Embodiment Robot Manipulation via a Unified Hand Action Space · https://arxiv.org/abs/2607.03570
  • [x] 2026 · Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models · https://arxiv.org/abs/2606.11324
  • [x] 2026 · The Embodiment Gap in Robot Foundation Models · https://arxiv.org/abs/2608.18433
  • [x] 2026 · Ambient Diffusion Policy: Imitation Learning from Suboptimal Data in Robotics · https://arxiv.org/abs/2606.12365
  • [x] 2026 · From Foundation to Application: Improving VLA Models in Practice · https://arxiv.org/abs/2607.06403
  • [x] 2026 · Keypose Exploration: Efficient Automatic Trajectory Labelling and Cross-Embodiment Policy Transfer · https://arxiv.org/abs/2606.29028
  • [x] 2026 · Auditing Instruction-Trajectory Mismatches in Multimodal Robot Demonstrations · https://arxiv.org/abs/2608.07895
  • [x] 2026 · Geometric Entropy: When Trajectory Diversity Helps and Hurts in Imitation Learning · https://arxiv.org/abs/2606.20871
  • [x] 2026 · Physics Filtering Favors the Generalization of Robot Learning · https://arxiv.org/abs/2608.22701
  • [x] 2026 · Set-Supervised Diffusion Policy: Learning Action-Chunking Diffusion through Corrections · https://arxiv.org/abs/2606.01865
  • [x] 2026 · VirTooS: A ROS 2 - Unity Virtualization Toolkit for Fleet Management of Autonomous Mobile Robots · https://arxiv.org/abs/2608.26066 （归位跳过：非数据×训练方向）
  • [x] 2026 · SUPER ODOMETRY 2.0: Resilient Odometry via Hierarchical Adaptation · https://arxiv.org/abs/2608.25427 （归位跳过：非数据×训练方向）
  • [x] 2026 · Robust Bimanual Vision-Language-Action Models via Embarrassingly Simple Modality Masking · https://arxiv.org/abs/2608.22419 （归位跳过：非数据×训练方向）
  • [x] 2026 · LD4WAM: Learning Latent Dynamics from Human Videos for World Action Models · https://arxiv.org/abs/2608.22403
  • [x] 2026 · Model-Based Reinforcement Learning for Heterogeneous Multi-Robot Task Assignment Under Distribution Shifts · https://arxiv.org/abs/2608.21554 （归位跳过：非数据×训练方向）
  • [x] 2026 · Model-Free Adaptive Parameter Tuning for Efficient Multi-Robot Warehouse Operations · https://arxiv.org/abs/2608.21533 （归位跳过：非数据×训练方向）
  • [x] 2026 · SRL-MPC: Shape-Aware Reinforcement Learned Model Predictive Control · https://arxiv.org/abs/2608.21175 （归位跳过：非数据×训练方向）
  • [x] 2026 · Rethinking Demonstration Unlearning in Imitation Learning for Robotics · https://arxiv.org/abs/2608.20784
  • [x] 2026 · RoboEdit: Turning Human Manipulation Videos into Scalable Robot Experience · https://arxiv.org/abs/2608.18948
  • [x] 2026 · PRISM: Precision and contact-rich Real-world Industrial Skill dataset with Multimodal sensing · https://arxiv.org/abs/2608.17962
  • [x] 2026 · NebulaVLA: A Dual-Frequency Vision-Language-Action Model With Guide Action for Robotic Manipulation · https://arxiv.org/abs/2608.16503 （归位跳过：非数据×训练方向）
  • [x] 2026 · Scaling Manual-Grounded Appliance Manipulation with Data Synthesis and Unified Planning · https://arxiv.org/abs/2608.15863
  • [x] 2026 · Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical Manipulation · https://arxiv.org/abs/2608.14379 （归位跳过：非数据×训练方向）
  • [x] 2026 · AdvDex: Learning Dexterous Manipulation from Human Demonstrations via Joint-Aligned Actions and Adversarial Learning · https://arxiv.org/abs/2608.14028
  • [x] 2026 · H2R-Bench: Benchmarking Human-to-Robot Manipulation Video Generation in World Models · https://arxiv.org/abs/2608.13049
  • [x] 2026 · Attune: A Self-Annotation Tool for Understanding Robot Operator Attention Profiles · https://arxiv.org/abs/2608.12650 （归位跳过：人因/监督界面注意力标注，非机器人训练数据引擎）
  • [x] 2026 · G0.5: One Autoregressive Stream for Robot Reasoning and Action · https://arxiv.org/abs/2608.11739 （归位跳过：非数据×训练方向）
  • [x] 2026 · RynnValue: Scaling Robotic Value Foundation Models with Temporal Distance · https://arxiv.org/abs/2608.09853 （归位跳过：非数据×训练方向）
  • [x] 2026 · JEPA-WAM: Learning Vision-Language-Action Policies with Joint-Embedding World Modeling · https://arxiv.org/abs/2608.09381 （归位跳过：非数据×训练方向）
  • [x] 2026 · Ego-OSCAR: Egocentric Open source Stereo CAptuRe System · https://arxiv.org/abs/2608.08285
  • [x] 2026 · Are Visual Place Recognition Models Recognizing Places or Conditions? Distractor-Augmented Evaluation and Condition Suppression · https://arxiv.org/abs/2608.06847 （归位跳过：非数据×训练方向）
  • [x] 2026 · Scalable Long-Horizon Planning with Staggered Updates for Lifelong MAPF · https://arxiv.org/abs/2608.06702 （归位跳过：非数据×训练方向）
  • [x] 2026 · CrossTracer: Cross-Embodiment Navigation via VLA Model Reasoning and Trace Residuals Adapting · https://arxiv.org/abs/2608.06688 （归位跳过：非数据×训练方向）
  • [x] 2026 · Anytime Global Tensor Motion Planning · https://arxiv.org/abs/2608.25830 （归位跳过：非数据×训练方向）
  • [x] 2026 · DESCENT: Directed Edge Scene Encoding for Airport Surface Movement Prediction · https://arxiv.org/abs/2608.26002 （归位跳过：非数据×训练方向）
  • [x] 2026 · VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning · https://arxiv.org/abs/2608.26105 （归位跳过：非数据×训练方向）
  • [x] 2026 · Performance-guided Task-specific Optimization for Multirotor Design · https://arxiv.org/abs/2510.04724 （归位跳过：非数据×训练方向）
  • [x] 2026 · Learning to Accelerate Vision-Language-Action Models through Adaptive Visual Token Caching · https://arxiv.org/abs/2602.00686 （归位跳过：非数据×训练方向）
  • [x] 2026 · RoboMME-Interference: Benchmarking Robot Memory Under Interference · https://arxiv.org/abs/2606.22338 （归位跳过：非数据×训练方向）
  • [x] 2026 · Model-Agnostic Open-Set Air-to-Air Visual Object Detection for Reliable UAV Perception · https://arxiv.org/abs/2509.09297 （归位跳过：非数据×训练方向）
  • [x] 2026 · Three-Way Open-Set Detection for Robust Autonomous Navigation · https://arxiv.org/abs/2511.15343 （归位跳过：非数据×训练方向）
  • [x] 2026 · Latent Chain-of-Thought World Modeling for End-to-End Driving · https://arxiv.org/abs/2512.10226 （归位跳过：非数据×训练方向）
  • [ ] 2026 · CLAP: Cross-Embodiment Video World Models are Zero-Shot Physical Simulators · https://arxiv.org/abs/2608.27406
  • [ ] 2026 · Marine Autonomous Vehicle Fleet Scheduling to Maximise Scientific Impact · https://arxiv.org/abs/2608.27271
  • [ ] 2026 · SpatialCrafter: Single Image World Modeling with Generative 3D Proxies · https://arxiv.org/abs/2608.27073
  • [ ] 2026 · 4DSynth: Controllable Procedural World Synthesis for Dynamic Embodied Simulation · https://arxiv.org/abs/2608.26947
  • [ ] 2026 · Beyond Shallow-Water Photorealism: Physically and Sensor-Grounded Simulation for Deep-Sea Robotics · https://arxiv.org/abs/2608.26888
  • [ ] 2026 · TrapVLA: Trapping Vision-Language-Action Models in Configured Failure Modes · https://arxiv.org/abs/2608.26578
  • [ ] 2026 · Closing the Loop on the Poppy Humanoid: Bipedal Locomotion with Linear-Quadratic Control and Learned Cost Functions · https://arxiv.org/abs/2608.26505
  • [ ] 2026 · Cross-Platform Benchmark of Neural 3D Reconstruction for Autonomous Laboratory Robots · https://arxiv.org/abs/2608.26383

已精读