ECCV2026 Robotics & Embodied AI AI paper notes paper summaries Multimodal/VLM Robotics Navigation Agents Reinforcement Learning Reasoning

🤖 Robotics & Embodied AI¶

🎞️ ECCV2026 · 129 paper notes

📌 Same area in other venues: 🧠 NeurIPS2026 (2) · 📷 CVPR2026 (146) · 🔬 ICLR2026 (162) · 💬 ACL2026 (11) · 🧪 ICML2026 (53) · 🤖 AAAI2026 (30)

🔥 Top topics: Multimodal/VLM ×44 · Robotics ×33 · Navigation ×20 · Agents ×6 · Reinforcement Learning ×6

VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Model: VLA-JEPA generates latent actions from a vision-language model that sees only initial images and instructions, learns manipulation semantics by predicting future V-JEPA2 states, and generates robot actions through flow matching; Tables 1 and 3 report success rates of 97.2% on LIBERO and 78.1% on LIBERO-Plus.
VLA-R1: Enhancing Reasoning in Vision-Language-Action Models: Built on Qwen2.5-VL-3B, VLA-R1 first learns geometrically grounded embodied reasoning from VLA-CoT-13K and then uses GRPO to optimize region, trajectory, and format rewards, reaching 36.51% IoU and 91.74 average trajectory error in-domain, although its real-robot results require caution about small evaluation samples and inconsistent reported numbers.
RoboClaw: An Agentic Framework for Scalable Long-Horizon Robotic Tasks: RoboClaw uses the same vision-language model (VLM) agent for robot data collection and execution, pairing forward actions with inverse reset behaviors and supervising multiple vision-language-action (VLA) policies online; real-robot experiments show reduced human collection time and improvements across four manipulation policies as closed-loop data accumulate.
Robobench: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models as Embodied Brain: RoboBench evaluates 18 models using 6092 questions across five embodied cognitive dimensions and a DAG-guided MLLM world-simulator judge, showing that strong general question answering and verbal planning do not guarantee implicit-intent understanding, spatial grounding, or real execution failure diagnosis.
EchoVLA: Robotic Vision-Language-Action Model with Synergistic Declarative Memory for Mobile Manipulation: EchoVLA separately maintains voxel scene memory for what exists where and episodic memory for recent interactions, retrieves both at different granularities to condition base and arm diffusion policies, and achieves 0.31 mean success on RoboCasa mobile manipulation versus 0.20 for π0.5 in the same table.
World Models for Learning Dexterous Hand-Object Interactions from Human Videos: DexWM conditions a latent world model on hand-keypoint differences and camera motion, using hand-consistency supervision to preserve local detail; after pretraining on human and parallel-jaw robot videos and fine-tuning on about 4 hours of target-robot simulation exploration, model-based planning achieves 10/12 successful real-robot grasps, where zero-shot means unseen test skills on the target embodiment and no real-robot fine-tuning—not the absence of all target-embodiment data.
RL from Physical Feedback: Aligning Large Motion Models with Humanoid Control: RLPF fine-tunes a large motion model with both simulated humanoid tracking success and text-motion semantic correspondence, while Hierarchical RLPF alternately improves the generator and controller to make motions executable without collapsing into instruction-independent standing.
Scaling Verification Can Be More Effective than Scaling Policy Learning for Vision-Language-Action Alignment: Without changing the deployed policy's parameters, CoVer-VLA uses a contrastive verifier trained on successful demonstrations to jointly select instruction rephrases and action chunks, raising average task progress on PolaRiS from 40.0% to 53.9% and success rate from 3.8% to 13.1%, showing that verification offers an effective alternative to further policy training, subject to candidate quality and additional compute costs.
MobileVLA-R1: Reinforcing Vision-Language-Action for Mobile Robots: MobileVLA-R1 establishes a parseable reasoning-and-action interface through multi-granularity chain-of-thought supervision, then uses offline GRPO rewards for continuous movement direction, discrete behaviors, and output format, reaching 68.3% SR on R2R-CE Val-Unseen and demonstrating quadruped closed-loop execution through fixed onboard/remote hybrid deployment.
Towards Generalizable Robotic Manipulation in Dynamic Environments: To address the limitation that existing VLAs can only grasp static objects and fail when facing moving targets, this work constructs DOMINO, a dynamic manipulation benchmark containing 117k trajectories across 35 tasks. Furthermore, it proposes the PUMA architecture, which equips VLAs with "historical optical flow + object-level future feature prediction," achieving a 6.3% absolute success rate increase on dynamic tasks.
Mechanistic Finetuning of Vision-Language-Action Models via Few-Shot Demonstrations: Robotic Steering uses 20 robot demonstrations to identify attention heads that best predict actions and applies head-selective LoRA fine-tuning, outperforming full-head LoRA on most real-robot tasks while better retaining robustness and unseen-task capabilities, although not every task benefits.
S-VAM: Shortcut Video-Action Model by Self-Distilling Geometric and Semantic Foresight: S-VAM self-distills geometric and semantic foresight from future videos generated along the same diffusion trajectory into one-step video features, then trains an action expert that also retains the original features, reaching 4.16 average completed tasks on CALVIN and 72.8% average success on MetaWorld.
FindingDory: A Benchmark to Evaluate Memory in Embodied Agents: FindingDory uses 60 task templates in Habitat to turn object-rearrangement histories into memory-dependent navigation instructions, with hierarchical metrics showing that even GPT-4o reaches only 27.3% high-level success among frozen vision-language models (VLMs), while remembering the right target still does not guarantee efficient navigation.
EvoVLA: Self-Evolving Vision-Language-Action Model: EvoVLA combines stage-aligned rewards, pose-based exploration, and selective memory on OpenVLA-OFT to reduce states that look complete but are physically incomplete, improving Discoverse-L average success from 59.0% to 69.2% and reducing hallucination rate from 38.5% to 14.8%.
RoboMirror: Understand Before You Imitate for Video to Humanoid Locomotion: RoboMirror generatively reconstructs motion latents from video semantics and feeds them to a physics-distilled diffusion policy, reducing total pipeline time from 9.22 seconds to 1.84 seconds while increasing success from 0.94 to 0.95 against the pose-driven approach on Motion-X.
Towards Unified World Models for Visual Navigation via Memory-Augmented Planning and Foresight: UniWM alternates action and next-frame prediction within one multimodal autoregressive model, using intra-step and cross-step KV memory to constrain successive rollouts; it raises Go Stanford success rate from NWM's 0.45 to 0.75 and reaches 0.42 on unseen TartanDrive.
Persistent Robot World Models: Stabilizing Multi-Step Rollouts via Reinforcement Learning: PersistWorld branches multiple future videos from its own increasingly imperfect histories and uses multi-view visual rewards for contrastive denoising post-training, reducing external-view LPIPS from 0.081 to 0.070 against reproduced Ctrl-World on 14-step DROID rollouts and receiving approximately 80% preference in a blind study.
Restoring Linguistic Grounding in VLA Models via Train-Free Attention Recalibration: ICBench exposes linguistic blindness when VLAs complete the original task despite scene-inconsistent instructions, and train-free attention recalibration with IGAR reduces this erroneous execution, lowering π0's success rate on LIBERO-Goal spatial contradictions from 76.4% to 36.4%, although improvements do not extend to every model and condition.
Habitat-GS: A High-Fidelity Navigation Simulator with Dynamic Gaussian Splatting: Habitat-GS extends Habitat-Sim with zero-copy Gaussian rendering, offline-prepared drivable avatars, and dynamic capsule obstacles while retaining its navigation interfaces; under a fixed budget, mixed mesh/GS training raises GS-test success from 61.30% to 79.60%, although real-world transfer evidence remains limited to open-loop video evaluation.
OmniStream: Mastering Perception, Reconstruction and Action in Continuous Streams: OmniStream converts the DINOv3 image encoder into a causal video backbone with a historical KV-cache, jointly trains it with self-supervised representation learning, geometric reconstruction, and language alignment, and reuses the same frozen visual features for perception, spatial reasoning, and robotic manipulation, reaching 3.885 average completed sequence length on CALVIN ABC-D.
RAE-NWM: Navigation World Model in Dense Visual Representation Space: RAE-NWM moves action-conditioned navigation prediction into frozen DINOv2 space, using CDiT-DH and generation-time-dependent action gating to improve long-horizon structural stability and reach 78.95% success in Habitat, with remaining trade-offs in path efficiency and texture-sensitive environments.
Affogato: Open-Vocabulary Affordance Grounding with Automated Data Generation at Scale: Affogato establishes a fully automated annotation pipeline by cascading three foundation models (Gemma3, Molmo, and MobileSAM), generating a scale of 750K open-vocabulary affordance heatmap dataset (Affogato-750K) from 150K 3D assets in Objaverse. It designs a unified cross-modal architecture, Espresso-3D/2D, to validate the effectiveness of this dataset as a general supervision signal—demonstrating consistent improvements across 3D and 2D affordance grounding benchmarks after pre-training, with the largest gains observed on unseen categories.
Diffusion Models are Open-World Affordance Learners: Leveraging Generative Priors for 3D Affordance Learning: DAG uses a frozen text-to-image diffusion model to extract human-object interaction features and combines them with action text and point-cloud geometry to predict 3D affordance masks, reaching 16.09 mIoU against GREAT's 12.05 on PIAD2 unseen affordances, although its few-shot results do not establish superiority over all fully trained baselines.
One-Step Flow Policy: Self-Distillation for Fast Visuomotor Policies: OFP combines interval self-consistency, self-guided regularization, and action warm-start to generate robotic action chunks with one network evaluation, achieving 71.6% average success across 56 simulated 3D manipulation tasks versus 66.4% for 100-step DP3, while reducing action-chunk generation time from 3225.67 ms to 17.58 ms.
Transport Discrepancy as a Reliability Signal for Vision-Language-Action Models: DiG uses sliced Wasserstein transport cost between observation features and action representations as an internal reliability signal to gate feature residuals, training sample weights, and inference-time refinement, improving \(\pi_{0.5}\) from 92.4% to 96.4% on LIBERO-Long and from 41.4% to 52.6% on low-data RoboCasa.
AgentVLN: Towards Agentic Vision-and-Language Navigation: AgentVLN uses Qwen2.5-VL-3B to schedule perception and planning skills through geometrically grounded image prompts, with local correction and on-demand depth queries resolving navigation ambiguities, achieving 67.2% / 69.5% SR on R2R-CE / RxR-CE Val-Unseen.
CoMaTrack: Competitive Multi-Agent Game-Theoretic Tracking with Vision-Language-Action Models: CoMaTrack trains two VLA robots to compete for following the same language-specified target, using asymmetric rewards and SFT-anchored GRPO as a dynamic curriculum; its 3B tracker reaches 92.1%, 74.2%, and 57.5% success on EVT-Bench STT, DT, and AT, with same-model STT success increasing from 89.5% under single-agent RL to 92.1%.
Dual-Anchoring: Addressing State Drift in Vision-Language Navigation: Dual-Anchoring adds training supervision for completed instruction segments and recently passed landmarks to streaming vision-language navigation, raising unseen R2R-CE SR from StreamVLN's 56.9% to 65.6% without retaining the auxiliary reconstruction branch in default deployment.
From Passive Observer to Active Critic: Reinforcement Learning Elicits Process Reasoning for Robotic Manipulation: PRIMO R1 combines initial images, process videos, and current images with a structured chain-of-thought cold start and GRPO driven by final progress error to train a 7B robotic process evaluator, achieving average MRA 82.90 and MAE 15.52 in a four-environment comparison and 67.0% accuracy on the unseen RoboFail benchmark.
GEM: Generative Supervision Helps Embodied Intelligence: GEM adds conditional depth generation to embodied vision-language model training and progressively aligns its components to encode geometry in shared representations, raising the 8B model's VSI-Bench score from 68.6 with same-data SFT to 70.6, while its GEM-VLA extension reaches a 96.1% average success rate on LIBERO.
MM-Nav: Multi-View VLA Model for Robust Visual Navigation via Multi-Expert Learning: MM-Nav combines the skills of depth-based RL experts for reaching, squeezing, and dynamic obstacle avoidance into a four-view RGB 7B vision-language-action (VLA) model through real-world VQA co-training and capability-balanced online distillation, achieving an average success rate of 88.1% on InternVLA-N1 System-1 and approximately 7 Hz continuous velocity prediction in server-assisted real-world deployment.
SPEAR: A Simulator for Photorealistic Embodied AI Research: SPEAR combines Unreal Engine runtime reflection, defined within-frame execution order, and NumPy shared-memory interoperability into a general Python simulation interface, exposing 14,485 underlying UE functions and delivering 1920x1080 images to Python at 73.4 FPS with two frames of rendering latency.
EgoCogNav: Cognition-aware Human Egocentric Navigation: EgoCogNav treats perceived path uncertainty as a supervised internal state that conditions joint trajectory and head-motion forecasting, reducing ADE/FDE by approximately 3.8%/5.0% against an adapted EgoCast on the CEN held-out test set while achieving a Spearman correlation of 0.788 for uncertainty prediction.
HiPolicy: Hierarchical Multi-Frequency Action Chunking for Policy Learning: HiPolicy makes a diffusion policy jointly predict low-frequency chunks covering longer horizons and high-frequency chunks for precise control, then adaptively selects an execution branch using sampled action entropy; the main table reports overall average success on selected RoboTwin tasks rising from 37% to 60% for DP and from 41% to 59% for DP3, but these aggregates cannot all be reproduced by simply averaging the visible task rows, and adaptive acceleration is evaluated separately.
HomeGuard: VLM-based Embodied Safeguard for Identifying Contextual Risk in Household Task: HomeGuard trains an independent safeguard with context-guided reasoning that localizes interaction targets and environmental constraints before judging risks, raising the 8B model's risk match rate on HomeSafe-Bench from 33.20% to 74.90% while reducing false alarms on safe scenarios from 32.62% to 13.14%.
HUGE-Bench: A Benchmark for High-Level UAV Vision-Language-Action Tasks: HUGE-Bench uses aligned 3DGS-Mesh digital twins reconstructed from real scenes to evaluate short-instruction, multi-stage UAV tasks jointly for process completion and collision safety; across 4 scenes, 8 tasks, and 2.56 million meters of trajectories, the strongest baseline, π0.5, still achieves only 0.581 average trajectory coverage.
RoboStream: Weaving Spatio-Temporal Reasoning with Memory in Vision-Language Models for Robotics: RoboStream binds object-level visual evidence to 3D geometry through STF-Tokens and continuously records action-induced state changes in a CSTG, enabling a vision-language model without task-specific fine-tuning to check manipulation preconditions, track occluded objects, and achieve 90.5% average success on eight RLBench long-horizon tasks.
E3VS-Bench: A Benchmark for Viewpoint-Dependent Active Perception in 3D Gaussian Splatting Scenes: E3VS-Bench turns active perception into a quantifiable task: over 99 3D Gaussian Splatting scenes it builds 2,014 question-driven episodes that can only be answered through 5-DoF viewpoint control, enforces that they cannot be solved from prior knowledge or the initial view (unanswerable viewpoints become starts, answerable ones become goals, plus blind and start-view filtering), and evaluates agents with a VLM judge score (1–5), average steps, and collision rate — the best model reaches only 2.79 versus 3.53 for humans.
SAFE-Pruner: Semantic Attention–Guided Future-Aware Token Pruning for Efficient Vision-Language-Action Manipulation: SAFE-Pruner observes that VLA models attend to the same semantic entities across control timesteps (semantic attention consistency), and therefore uses the per-layer attention saliency cached at the most recent key timestep to forecast the deep-layer saliency that is not yet computable for the current frame; the forecast is fused with the current shallow-layer saliency for top-K token pruning, and a shallow-attention drift detector adaptively refreshes the key timestep, yielding up to 1.89x speedup with less than 1.5% drop in success rate without any training or architectural change.
Tactile Modality Fusion for Vision-Language-Action Models: TacFiLM turns a pretrained tactile encoder's embedding into per-channel FiLM scale/shift parameters that modulate the intermediate visual features of OpenVLA-OFT — without adding a single token to the language model — raising the in-distribution average success rate from 58.10% to 86.67% on real-robot peg insertion, USB/HDMI plugging and drawer-opening tasks while cutting peak contact force to roughly one third.
Text-Guided 6D Object Pose Rearrangement via Closed-Loop VLM Agents: A single pretrained vision-language model (VLM) alternates between evaluating a scene and proposing incremental object pose updates, corrects errors through re-rendered feedback, and combines three geometric prompting techniques to improve overall Open6DOR V2 6-DoF goal-pose prediction success from SoFar's 26.3% to 46.4% without fine-tuning.
UniBYD: A Unified Framework for Learning Robotic Manipulation Across Embodiments Beyond Imitation of Human Demonstrations: UniBYD aligns the state and action spaces of 2-, 3-, and 5-fingered robotic hands with a unified morphological representation (UMR), then uses a dynamic PPO steered by a "shadow engine" that anneals the dense imitation reward into a sparse goal reward, so policies trained from human demonstrations generalize across hand morphologies and beat ManipTrans by 44.08% on average in success rate on the self-built UniManip benchmark (+59.23 points on 5-fingered unimanual, +28.92 points on 5-fingered bimanual).
What Matters in RL-Based Methods for Object-Goal Navigation? An Empirical Study and A Unified Framework: This paper decomposes modular RL-based object-goal navigation (ObjectNav) systems into perception, policy, and test-time enhancement modules and runs one-factor-at-a-time replacements under a single Habitat protocol; it finds that upgrading perception or adding an inference-time strategy often yields larger gains than upgrading the policy network, and assembles those findings into a system that surpasses all prior methods on Gibson (SR 84.4%, SPL 48.2%), alongside a 98% human-expert baseline.
ESCAPE: Episodic Spatial Memory and Adaptive Execution Policy for Long-Horizon Mobile Manipulation: ESCAPE uses persistent spatial memory for both navigation and interaction-mask generation, while allowing a local manipulability monitor to preempt global routes, achieving 65.09%/60.79% test-seen/test-unseen success on ALFRED with step-by-step instructions and substantially improving path-length-weighted metrics.
KineBench: Benchmarking Embodied World Models via IDM-Free Kinematic Grounding: KineBench is an IDM-free closed-loop benchmark for embodied world models (EWMs): an explicit kinematic grounding pipeline (instance segmentation → monocular metric depth → CAD-constrained 6D pose tracking) recovers end-effector poses from generated videos and injects them into ManiSkill3 for physical execution, while two classical robot-centric metrics, SPARC and the Maruyama Manipulability Index, provide robot-side diagnostic signals; 7 frontier video models are evaluated over 20 manipulation tasks in four progressive suites.
Let's Reward Step-by-Step: Step-Aware Contrastive Alignment for Vision-Language Navigation in Continuous Environments: SACA uses frozen vision and language models to identify useful prefixes and divergence points in navigation trajectories, then repairs or rescues failures depending on whether the sampled group contains a success, reaching 60.3% / 60.3% success on R2R-CE / RxR-CE unseen validation without extra data or inference-time auditor overhead.
LIBERO-Safety: A Comprehensive Benchmark for Physical and Semantic Safety in Vision-Language-Action Models: This paper proposes LIBERO-Safety, a comprehensive benchmark for systematically evaluating the physical and semantic safety of Vision-Language-Action (VLA) models. It features a programmatic environment definition language (UBDDL) to generate safety-critical scenarios, an efficient keypose-driven trajectory synthesis pipeline yielding massive, collision-free demonstration data (19,664 trajectories), and a five-dimensional safety task taxonomy (75 tasks). The study reveals a fundamental tension in current VLA models between generalization capability and safe execution—while high-diversity training improves safety trajectory quality, task success rates remain bottlenecked by sub-optimal trajectory synthesis and semantic misalignment.
NavWM: A Unified Navigation World Model for Foresight-Driven Planning: NavWM proposes a unified navigation world model that jointly learns latent world reasoning (depth/semantic scene abstraction), anchor-based multimodal trajectory prediction, and Flow Matching conditional visual generation on a shared bidirectional Mamba backbone. This allows the world model itself to act as a closed-loop "visual foresight" planner—simulating future observations for each candidate trajectory and selecting the optimal path—achieving state-of-the-art (SOTA) performance in offline generation quality (PSNR 14.17 \(\rightarrow\) 17.34) and zero-shot navigation success rate (44%) across five robotic datasets.
R3DP: Real-Time 3D-Aware Policy for Embodied Manipulation: R3DP runs expensive VGGT geometry processing only on sparse keyframes, updates its features with a lightweight temporal network, and fuses camera geometry, enabling an RGB diffusion policy to achieve 69.0% average success over ten RoboTwin tasks while reducing observation encoding latency from 73.1 ms to 40.3 ms under a different operating configuration.
Understanding the Impact of Geometric Foundation Models on Vision-Language-Action Models: Using GR00T-N1.5 and VGGT as a fixed pair, this paper quantifies for the first time the "geometric gap" between a VLA and a geometric foundation model (GFM) via linear probing, distills the disparate geometry-injection schemes in the literature into three controlled, comparable routes—Early Fusion, Late Fusion, and Spatial Forcing—and systematically scans non-architectural factors such as training-data size, camera count, and reconstruction quality: under task-level fine-tuning geometric VLAs are not significantly better, and only when training data is scaled up or the setup degrades to a single camera does the value of geometric information become visible.
Beyond Dense Futures: World Models as Structured Planners for Robotic Manipulation: StructVLA trains a world model to predict sparse images selected around gripper transitions and motion pauses, then transfers this task-progress representation to action generation, achieving average success rates of 75.0% on SimplerEnv-WidowX and 94.8% on LIBERO.
CoToGrasp: Contact-Topology-Conditioned Dexterous Grasp Synthesis via Canonical Workspace Learning: CoToGrasp learns requested contact topologies using robot-hand geometry alone, transfers them to unseen objects through a canonical workspace, and validates and optimizes the resulting grasps, improving topology-averaged success from 21.13% to 26.72% and topology compliance from 14.28% to 17.18% on DexGraspNet without universally outperforming unconditioned planners in physical success.
Deform360: A Massive Multi-view Visuotactile Dataset for Deformable World Models: Deform360 records 1,980 interactions with 198 deformable objects using 41 views and bimanual tactile sensing, builds annotations by decoupling geometry from motion, and compares video and particle world models: explicit physical priors help low-data prediction, while large-scale pretraining supports visual generalization to new objects without guaranteeing action compliance.
DVG-WM: Disentangled Video Generation Enables Efficient Embodied World Model for Robotic Manipulation: DVG-WM explicitly splits a video world model into a low-resolution preview stage that learns dynamics and a high-resolution refinement stage that restores detail, using flow matching to map preview latents directly to high-resolution video latents; on LIBERO it improves video quality (FVD 152.36 vs 171.24 for CogVideoX-5B), reaches up to 3.97x inference speedup with only 4 refinement steps, and attains 62.2% average success across nine real-world tasks (vs 34.0% for a CogVideoX world model).
E-VLA: Event-Augmented Vision-Language-Action Model for Dark and Blurred Scenes: E-VLA transforms event camera streams into frame-like accumulated representations through a recent event count window. It then integrates these representations into the SmolVLA visual encoding pipeline via two lightweight approaches: zero-parameter overlay or a 13M hierarchical adapter. This enables the robot's Pick-and-Place success rate to jump from 0% to 90% under extremely low illumination of 20 lux, recovery from 0% to 25% under severe motion blur of 1000 ms, and maintaining 45% success under 20 lux in OOD generalization tests training only on 200 lux data.
Efficient Camera Pose Augmentation for View Generalization in Robotic Policy Learning: GenSplat reconstructs 3D Gaussian scenes from sparse, uncalibrated robot observations in a forward pass, stabilizes their geometry through 3D-prior distillation, and renders new-view demonstrations with unchanged action labels, increasing average \(\pi_0\) success from 48.9% to 62.2% under large camera perturbations at 0.14 seconds per synthesized frame.
Humanoid Whole-Body Manipulation via Active Spatial Brain and Generalizable Action Cerebellum: An Active Spatial Brain decides where to observe next and what to execute, while a Generalizable Action Cerebellum translates those decisions into navigation, body repositioning, and dexterous manipulation without task-specific real-robot demonstrations, achieving 60.0% success on hard obstacle avoidance versus 30.0% for CaP.
PolicyTrim: Boosting Intrinsic Policy Efficiency of Vision-Language-Action Models: PolicyTrim proposes a two-stage GRPO post-training framework. By utilizing dynamic execution window exploration to expand reliable action chunk lengths and incorporating step-saving rewards alongside group-anchoring stability regularization to eliminate redundant physical steps, it improves the end-to-end deployment speed of VLA robots by up to 5.83x without altering the architecture or requiring additional demonstration data.
Pondering the Way: Spatial-perceiving World Action Model for Embodied Navigation: SWAM proposes to combine trajectory planning and future observation generation in visual navigation, generating intermediate RGB-D frame sequences and corresponding action trajectories directly from start/goal RGB images in a single diffusion inference. This eliminates the traditional two-stage "sample-then-verify" paradigm and comprehensively outperforms baselines in terms of accuracy, efficiency, and zero-shot generalization.
Revisiting Parameter Redundancy in Vision-Language-Action Models: Insights from VLM-to-VLA Adaptation: This paper revisits the parameter redundancy problem in Vision-Language-Action (VLA) models, finding that the existing "prune-then-fine-tune" paradigm masks the mistaken deletion of critical parameters. The authors propose using the weight divergence (\(\Delta W\)) during the VLM-to-VLA adaptation process as a redundancy metric. Through controlled pruning diagnostic experiments, they reveal that the \(\Delta W\) signals exhibit strong heterogeneity across different modules. Based on these findings, they design a multi-module joint pruning scheme that achieves a 12%--30% parameter reduction on OpenVLA and \(\pi_{0.5}\) while maintaining approximately 90% of the original performance without any post-pruning recovery.
Tac2Real: Reliable and GPU Visuotactile Simulation for Online Reinforcement Learning and Zero-shot Real-World Deployment: Tac2Real pairs a frictional PNCG-IPC soft-body solver (implemented in Taichi, pluggable into Isaac Lab / MuJoCo and other engines) that emits 7×9×2 marker displacement fields at 4,465 FPS over 4,096 environments on multiple nodes and GPUs with a four-stage calibration pipeline, TacAlign, that closes the sim-to-real gap; a policy trained entirely in simulation transfers zero-shot to a real Franka, succeeding in 55 of 60 blind peg-insertion trials (91.7%).
Task-driven Processing with Coarse-to-Fine Glimpse-based Active Perception: CF-GAP is a task-driven perception front-end: it convolves multi-view examples of the search target over a downscaled scene to obtain a coarse search heatmap, picks 30 coarse glimpse locations with winner-takes-all plus inhibition-of-return, then drives a log-polar sensor in a closed loop on each coarse location so that the 2D centroid of a fine search map walks the fixation onto the target; only a handful of high-resolution RoIs and the glimpse locations are handed to an off-the-shelf instance detector, raising the AP of several state-of-the-art detectors by up to about 20 points on HR-InsDet and Robotools and letting lightweight detectors surpass their heavier counterparts.
VLA-Hijack: A Transferable Patch Attack against Vision-Language-Action Models via Visual Proprioception Hijacking: VLA-Hijack investigates visual self-localization representations as a robustness weakness in vision-language-action models, combining real-arm feature suppression with surrogate embodiment injection to reach 61.64% average task failure across 11 black-box LIBERO transfer models, with substantial architecture-dependent variation.
360CityArena: A Realistic Virtual Urban Navigation Benchmark for Embodied Agents: 360CityArena places 175 manually constructed tasks in an Akihabara street-network pose graph built from real panoramic videos, diagnosing seven urban navigation capabilities; the best evaluated model, Gemini-2.5 Flash, scores only 17.1% overall against 77.3% for locally familiar humans, exposing intertwined weaknesses in map-to-view alignment, sustained exploration, and fine-grained perception.
3DWay: Generalizing Robot Manipulation via 3D Consistent Waypoints: 3DWay expresses 3D trajectory generation as multi-view 2D coordinate prediction in a VLM's text interface, reconstructs executable waypoints using calibrated geometry, reaches 64.0% success on unseen RLBench tasks with fixed top-down control, and improves few-shot pi0 adaptation through local spatial guidance.
A scalar per patch from pre-trained ViTs enables fast moving navigation in the real world: This is a large-scale empirical study conducting 966 physical robot navigation trials (accumulating 24 km of fast motion at 0.7 m/s) in a real office building. It systematically compares frozen visual encoders (such as DinoV2, DinoV3, VC-1, AM-Radio, and Dune) on pure RGB PointGoal navigation. The study proposes a minimalist projection layer named PAP (Pure Attention Projection), which compresses each image patch into a single scalar attention value fed to the downstream agent. This bottleneck preserves navigation performance while spurring the self-emergence of interpretable heatmaps representing navigable areas, obstacles, and affordances.
AdaDexGrasp: Adaptive Dexterous Grasping via 3D Visuo-Tactile Representation Fusion: AdaDexGrasp predicts which hand part should contact each object region to generate an initial grasp, then maps measured touch back into 3D geometry to assess and correct unstable configurations, raising simulated success on unseen categories from 59% without adaptation to 83%.
Attention-DP3: Spatially Object-aware 3D Diffusion Policy via Geometry-aligned Attentional Conditioning: Attention-DP3 projects language-specified object masks onto observed point clouds and conditions an otherwise unchanged DP3 action denoiser with three complementary semantic fields, raising MetaWorld average success from 0.669 to 0.726 and improving real-robot manipulation under clutter.
AutoSpeed: Annotation-Free Stage-Adaptive Motion Speed Learning for Robot Manipulation: AutoSpeed proposes a model-agnostic training framework that learns stage-adaptive motion speeds without requiring any stage or speed annotations. By treating future trajectories at different speeds as candidate optimization targets and utilizing a composite cost function (prediction error vs. prediction horizon) to select the optimal candidate, the framework enables visuomotor policies to end-to-end learn to accelerate in simple phases and decelerate in fine-grained phases, simultaneously improving both success rates and execution efficiency across multiple simulation and real-world robot tasks.
Beyond Description: Cognitively Benchmarking Fine-Grained Action for Embodied Agents: CFG-Bench decomposes action understanding into physical interaction, temporal-causal relations, intention, and evaluation, using hybrid QA and counterfactual gating to expose the gap between description and execution knowledge; fine-grained data SFT raises Qwen2.5-VL-7B's average EB-ALFRED result from 4.7 to 9.7.
ComplexMimic: Human–Scene Interaction Imitation in Complex 3D Environments: ComplexMimic trains a strict motion-tracking expert and a collision-adaptive interaction expert, then routes supervision by motion regime and prioritizes difficult trajectories that still improve, distilling a single controller with 90.6% success on TRUMANS while retaining cross-domain pose-error trade-offs.
Depth-guided Multi-view Exposure Bracketing for HDR Robot Vision: DMEB distributes different exposures across synchronized cameras and uses external depth to align and fuse complementary observations for single-shot HDR reconstruction, achieving 39.40 dB PSNR-µ with three cameras on real modest-DR data while introducing evaluation data covering 121 real scenes and 20 synthetic videos.
Domain Adaptation with Adaptive Imagination for Visual Reinforcement Learning under Limited Target Data: AIDA maps target images into the source state space, rolls out a frozen policy and dynamics model, uses a three-way discriminator to cut off each trajectory once it leaves the target manifold, and applies state→image→state consistency to the retained prefix; with only 50 target trajectories, it obtains the best return ratio on 6 of 7 tasks and the lowest state RMSE on all seven.
EgoPHI: Estimating 3D Hand-Object Contact and Force from Egocentric Vision: EgoPHI is the first vision-based model that jointly predicts dense per-vertex 3D contact regions and interaction forces on both hands and articulated object meshes from a single egocentric RGB image and object template geometry, leveraging physics-based simulation supervision and real-world FTIR sensor validation.
Grasp-Oriented Non-Prehensile Manipulation via Learning a Graspability Field: GOMP reformulates preparatory pushing, sliding, and reorientation from reaching a specified object pose to reducing distance to a set of graspable configurations, then uses predicted distance to trigger grasping; it completes 31 of 50 end-to-end real-robot trials from challenging initial states, rather than achieving 97% overall success.
Gripper-aware Vision Language Action Models: Addressing the pervasive "gripper invariance" assumption in vision-language-action (VLA) models, this paper introduces MiGA, a comprehensive dataset of 103,000 trajectories across 5 distinct gripper types, and proposes GVLA, which combines multi-granularity soft prompt tokenization with a dual Mixture-of-Adapters to achieve strategy-level gripper conditioning and robust cross-embodiment adaptation.
Hypothesis Graph Refinement: Hypothesis-Driven Exploration with Cascade Error Correction for Embodied Navigation: This paper proposes Hypothesis Graph Refinement (HGR), which represents frontier predictions as revisable hypothesis nodes in a dependency-aware graph memory and couples goal-directed semantic exploration with verification-driven cascade correction, fundamentally resolving structural error accumulation in long-horizon embodied navigation.
MA-VLA: Multi-Arm Vision-Language-Action Model for Collaboration and Compositional Generalization: To overcome the severe arm-identity bias and static instruction limitations in multi-arm manipulation, MA-VLA decomposes complex instructions into per-arm atomic prompts and incorporates training-time Arm Shuffle, enabling robust zero-shot multi-arm compositional generalization across unseen collaboration topologies.
NutriBench-Kitchen: Benchmarking Embodied AI for Nutrition Management: This paper introduces NutriBench-Kitchen, a benchmark covering 160 long cooking videos and 1,500 human-verified multi-task QA pairs to evaluate Embodied Nutrition Management, and proposes Nutri-Vgent, an agent featuring decoupled episodic, food-state, and recipe memories that substantially bridges the performance gap in long-horizon state maintenance and constraint-aware planning.
One Demonstration Is Enough for Real-World Robotic Reinforcement Learning: AutoSERL replaces expensive human teleoperation and multi-demonstration dependence in real-world robotic reinforcement learning by deriving automated sliding-window guidance, stagnation recovery replays, and adaptive termination criteria from a single demonstration, achieving 100% success across 6 contact-intensive physical tasks while matching HIL-SERL.
Panoramic Affordance Prediction: Addressing the narrow Field-of-View and fragmented observations inherent to pinhole cameras in embodied AI, this paper pioneers the task of Panoramic Affordance Prediction, introduces PAP-12K—the first benchmark featuring 1,003 native 12K panoramic images and over 13k QA pairs—and proposes PAP, a training-free, human fovea-inspired pipeline that robustly handles severe geometric distortion, boundary discontinuity, and extreme scale variations.
Path-level Hindsight Instructions for Semantic Exploration in Vision-Language Navigation: Addressing the semantic instruction mismatch caused by exploratory deviations in Vision-Language Navigation (VLN), this paper introduces Φ-Nav, a unified on-policy framework that synthesizes path-level hindsight instructions via an LVLM conditioned on expert exemplars, filters hallucinations using landmark-aware trajectory-instruction alignment weighting, and bootstraps policy training with reduced expert demonstrations.
PixelPilot: Scalable Vision-Language-Action Models for End-to-End Autonomous Driving: PixelPilot introduces a decoupled planning and lifting paradigm that shifts policy optimization entirely to sensor-agnostic 2D image space with deterministic 3D lifting at inference, eliminating sensor entanglement and ego-status shortcut learning while achieving state-of-the-art open-loop and closed-loop driving performance.
Pose Anything Anywhere: Model-free Object Poses from Arbitrary References: PANY proposes a unified model-free 6D pose estimation framework that jointly trains multi-view geometric reasoning and cross-view correspondence learning in an end-to-end manner. It can directly estimate the 6D pose of unseen objects from arbitrary numbers of sparse reference images (including optional pose-free auxiliary views) without requiring CAD models or pre-registration, outperforming prior methods by a large margin on benchmarks such as YCB-V and LM-O.
Predicting Consequences and Reinforcing Navigation Policies with Latent World Models: This paper introduces a compatibility prediction Latent World Model (LWM) for visual robot navigation that replaces heavy pixel/feature reconstruction with action-conditioned latent compatibility prediction, enabling counterfactual training, policy supervision from unlabeled videos, and in-imagination GRPO reinforcement without environment interaction.
RelAfford6D: Relational 6D Affordance Graphs for Constraint-Driven Robotic Manipulation: RelAfford6D proposes a training-free robotic manipulation framework centered on Relational 6D Affordance Graphs, transforming open-world language instructions into part-level SE(3) kinematic constraint manifolds and closed-loop visual replanning to achieve superior zero-shot manipulation on both articulated and rigid objects.
Self-Evolving Just-In-Time Memory for Proactive Embodied Safety: Addressing the dilemma where embodied agents frequently cause dynamic physical risks during benign-goal household execution, this paper proposes the Self-Evolving Just-In-Time Memory (JIT-Memory) framework, combining a topological belief graph, intent-triggered factual rules, procedural meta-skills, and an annotation-free test-time verification loop to achieve proactive risk mitigation without stalling task progress.
Teaching Vision-Language-Action Models What to See and Where to Look: DriveTeach-VLA introduces Driving-aware Vision Distillation (DVD) and 2D Trajectory-Guided Prompts (2D-TGP) to decouple and ground spatial attention on critical traffic objects and feasible driving corridors, achieving 90.4 PDMS on NAVSIM and setting new state-of-the-art trajectory prediction accuracy on nuScenes.
Trajectory-Level Continuous Action Representation for Robotic Manipulation: To resolve the rigid entanglement between control frequency and action representations in visuomotor systems, CAT encodes action trajectories within a fixed real-time interval into a constant set of continuous latent tokens with frequency-aware rotary position embeddings and asymmetric causal decoding, consistently outperforming timestep-level and discrete VQ baselines across simulation and real-world long-horizon manipulation tasks.
Trust Your Instincts: Confidence-Driven Test-Time RL for Vision-Language-Action Models: Addressing the heavy dependence of existing VLA reinforcement learning on external environmental success feedback, this paper demonstrates a strong positive correlation between generative confidence and physical execution success, and introduces T2VLA—an architecture-agnostic test-time RL framework that self-improves via dual-expert bootstrapping and DTW-based intrinsic rewards across discrete and continuous flow-based VLAs.
Unified Prediction and Planning via Conflict-Aware Disjoint Parameter Training: To tackle severe representational skill conflicts in compact shared encoders for robot navigation, this paper introduces Disjoint Parameter Training (DPT) and Sparse Merging, achieving safe ego planning and accurate neighbor prediction without any inference latency overhead.
AeroVLA: A Vision-Language-Action Model for UAV Navigation via Minimalist End-to-End Control: AeroVLA combines front/down views, coarse target bearings, and native numerical tokens in a reactive OpenVLA-7B policy that jointly controls UAV motion and termination, achieving 37.58% SR on unseen TravelUAV maps versus LongFly's 11.27%, while still relying on a target-bearing prior and a low-level flight controller.
AMCoNav: Asynchronous Multi-module Collaborative Framework for Embodied Visual Navigation: AMCoNav keeps a lightweight navigation policy running while large models reason on demand in the background, then uses shared context and probability fusion to choose exploration, verification, or direct navigation, achieving 60.0% / 60.0% / 53.3% success on the three HM3D-OVON splits and reducing runtime against a synchronous counterpart without guaranteeing shorter paths.
BeTTER: Diagnose the Illusion of Embodied Reasoning in Vision-Language-Action Models: BeTTER separates motor competence from semantic, compositional and state reasoning through task interventions designed to keep familiar motions executable: three representative VLAs achieve only 5%, 15% and 0% success on unseen subgoal compositions, while physical robot tests also reveal behavioral inertia and phase conflicts.
ChronoFlow Policy: Unifying Past-Future Interaction Flow in Visuomotor Policy Learning: ChronoFlow-Policy represents past, current, and future object-gripper motion as unified 3D keypoint trajectories, jointly trains future-flow recovery and action prediction, reaches average success rates of 72% on MetaWorld and 66% on RoboTwin 2.0, and improves final-stage real-world Swap-Easy success from 20% without history to 93% with history.
CulinaryCut: A Physics-aware Vision-Language-Action Benchmark for Food Cutting via Material Point Method: CulinaryCut couples MLS-MPM with ManiSkill to generate food-cutting demonstrations with contact-force records, exposing both ratio-grounding and physical-severing failures while improving RDT's orange-cutting success in IntSim from 23% to 59% without changing its architecture or supplying force inputs.
DiNBV-Grasp: Real-Time Distance-Aware Two-Stage Next-Best-View for Robotic Grasping: DiNBV-Grasp separates choosing a viewpoint that resolves occlusion from choosing an observation distance that supports reliable grasping, using a distilled semantic planner for the former and a success-probability scorer for the latter; its simulation plug-in raises ManipLLM Seen/Unseen success from 56.00%/51.00% to 65.62%/58.42% without modifying the underlying manipulation model.
Environmental Change Detection for Real-World Change Analysis: This paper introduces Environmental Change Detection (ECD), a practical paradigm that lifts the unrealistic assumptions of predefined and strictly aligned query-reference pairs by retrieving reference candidates from an uncurated database via Visual Place Recognition (VPR) and assembling a query-oriented representation via a multi-scale Environment Mosaic Composer (EMC) and semantic aggregator, outperforming state-of-the-art detectors in open, non-aligned real-world settings.
EvoWorld: A World-Model-Centric Framework for Continuous Self-Evolution of Modular Embodied Skills: EvoWorld decouples general physical dynamics from specialized embodied skills, utilizing an action-conditioned video world model for counterfactual rollout verification and a VLM diagnostic loop for modular, targeted policy self-evolution without full-system retraining.
Exo2EgoPolicy: Pose-Aligned Cross-View Policy Learning: Exo2EgoPolicy bypasses pixel-level video hallucination by formulating exocentric-to-egocentric translation directly as an SE(3) hand-object pose diffusion process, combining temporal regularization and object-relative frames to enable scalable robot policy co-training with 20–30% higher success rates.
FedNASP: Federated Vision-Language Navigation with Adaptive Step-Wise Personalization: To resolve the vulnerability of conventional federated averaging and static personalization under severe cross-environment non-IID shifts and time-varying decision uncertainty in Vision-Language Navigation (VLN), FedNASP introduces a lightweight Step-wise Personalization Modulator (SPM) that dynamically predicts step-wise modulation scalars from navigation state and history, combined with structure-aware personalized prefix injection (SPI) on critical attention blocks, achieving robust decentralized navigation without communicating local personalized weights.
Fixed Reality, Diffused Possibility: Disentangling Stochastic and Deterministic Latent for Cluttered Grasping: Addressing the entanglement of deterministic geometric structure and stochastic strategic uncertainty in heavily cluttered scenes, this paper proposes a split-latent hierarchical diffusion framework (FRDP) that decomposes scene representations into a fixed geometric latent and a diffused contextual latent, regularized by a differential-geometric Harmonic Grasp Field to produce smooth, obstacle-aware, and physically consistent 6-DoF grasps.
GameWorlds: Towards Standardized and Verifiable Evaluation of Multimodal Game Agents: GameWorld introduces the first standardized, state-verifiable benchmark for multimodal game agents across 34 browser games and 170 tasks, decoupling inference latency from execution via a paused sandbox and unifying computer-use and generalist interfaces to reveal profound capability gaps compared to human players.
Grounding Sim-to-Real Generalization in Dexterous Manipulation: An Empirical Study with Vision-Language-Action Models: Through over 10,000 physical robot trials across five bimanual manipulation tasks, this work systematically decouples and quantifies the critical factors governing zero-shot Sim-to-Real transfer for Vision-Language-Action (VLA) models, demonstrating that spatial variations outshine appearance factors and that frame-wise randomization combined with reinforcement fine-tuning dramatically boosts real-world robustness.
Guide, Think, Act: Interactive Embodied Reasoning for Vision-Language-Action Model: Addressing the fragility of conventional "Sense-to-Act" VLA policies and the uncorrectable nature of self-contained embodied CoT models, GTA-VLA introduces a unified "Guide, Think, Act" framework where sparse human spatial priors (affordance points, bounding boxes, motion traces) directly condition structured spatial-visual Chain-of-Thought reasoning, which is then asynchronously executed via a fast Flow-Matching action head to achieve 81.2% success on SimplerEnv WidowX and superior failure recovery under OOD visual and spatial ambiguities.
LaMP: Learning Vision-Language-Action Policies with 3D Scene Flow as Latent Motion Prior: LaMP proposes a dual-expert VLA framework that leverages a conditional flow-matching Motion Expert to generate one-step partially denoised 3D scene flow hidden states, injecting them into an Action Expert via gated cross-attention to substantially enhance physical grounding and OOD robustness without requiring inference-time depth sensors.
LARY: A Latent Action Representation Yielding Benchmark: Meituan introduces LARYBench, the first unified benchmark decoupling latent action representation evaluation from downstream policy learning across 1.2M video clips and 151 action classes, revealing that off-the-shelf general visual encoders without action supervision systematically outperform specialized embodied latent action models.
LEAP-VLA: Latent-Enhanced Action Prototyping via Continuous Residual Latent Spaces for Vision-Language-Action Models: Addressing the reliance of vision-language-action models on massive parameter counts and expensive iterative diffusion decoders, LEAP-VLA proposes a decoupled action representation paradigm: Stage 1 learns a structured continuous action manifold via Multi-level Soft Residual Quantization (MSRQ), and Stage 2 trains a lightweight VLM-external aligner to perform single-pass soft prototype selection, outperforming SOTA models of several times its size using only a 1B backbone.
Less is More: Reducing Complexity in Vision-Language-Action Systems: Targeting the fragmentation, benchmark-specific overfitting, and bloated engineering in Vision-Language-Action (VLA) systems, this paper introduces StarVLA-α, demonstrating that a strong pretrained VLM backbone paired with a lightweight MLP continuous action head and simple zero-padding matches or outperforms state-of-the-art complex architectures across multiple simulation benchmarks and real-world physical robot evaluations.
Lifting Ego World Models for Planning and Control: Addressing the challenge that high-dimensional joint action spaces make human-like world models intractable for search-based planning, this paper introduces a lightweight policy that maps sparse 2D leaf joint waypoints on the current frame into joint action sequences, lifting frozen world models into an 8D action space that reduces CEM planning error by 3.8× at substantially lower compute cost.
M2Tok: Multi-head Multi-codebook Discrete Action Tokenization for Vision-Language-Action Models: Addressing the discretization bottleneck caused by semantic entanglement and limited expressivity in existing VLA action tokenizers, M2Tok introduces orthogonal latent subspace decomposition and combinatorial multi-codebook quantization, reducing the action reconstruction L1 error to 0.0024 and boosting bimanual manipulation success rate to 51% on RoboTwin.
MuSix: Multi-scale Mixture of World Models for Embodied Agents in Evolving Environments: This paper proposes the MuSix framework, which explicitly decouples "scale determination" from "model selection" in world models via a two-stage scale-aware routing mechanism driven by experiential distance inspired by Construal Level Theory (CLT). It introduces scale-dependent forgetting rates and gated cross-scale knowledge transfer. This allows low-level immediate knowledge to refresh rapidly while high-level abstract knowledge remains stable, achieving multi-scale adaptive evolution for embodied agents in dynamic environments.
NeuIDO: Neural Intrinsic Dynamics Operator for Physics-Informed 4D World Models: NeuIDO casts physics-informed 4D world dynamics modeling as neural operator learning from video spaces to constitutive function spaces, leveraging a two-stage training scheme to achieve real-time zero-shot dynamics inference and efficient few-shot adaptation for novel scenes.
ORION: Ordinal Neural Collapse as a Representation Prior for Visual Navigation: Addressing ambiguous representation learning and severe intersection trajectory flickering caused by indirect supervision from single action losses in end-to-end visual imitation learning, ORION discretizes goal-relative headings into ordered control categories and imposes ordinal neural collapse geometry (supervised 1D projection alignment and orthogonal variance suppression) during encoder pretraining, boosting navigation success rate at multi-way intersections by up to 26 percentage points.
PARL-VLA: Pruning-Aware On-Policy Reinforcement Learning for Vision-Language-Action Model: Addressing the severe exploration-exploitation mismatch and training collapse caused by naive token pruning in on-policy VLA post-training, PARL-VLA records rollout pruning configurations and replays them during policy updates to align likelihood ratios, coupled with a spectrum pruning curriculum that discards 55% of visual tokens while accelerating training and boosting out-of-distribution robustness.
RAGrasp: A Retrieval-Augmented Framework with Diversity-Aware Modeling for Dexterous Grasp Generation: Addressing the twin challenges of grasp success rates being bottlenecked by random initial latent states and grasp diversity collapsing under greedy single-label classification, RAGrasp introduces a distribution-aligned geometry-aware retrieval initialization strategy alongside an exploration-plus-fallback dual-head grasp-type selector, achieving substantially improved grasp success rates while enabling flexible diversity tuning.
RoMan-4D: Learning Robot Arm Manipulation from 4D World Models: RoMan-4D (GEM-4D) introduces an asymmetric dual flow-matching framework that distills 4D geometry foundation model representations into a video DiT backbone during training with zero inference overhead, coupled with an adaptive confidence-gated inverse dynamics system to derive executable 6-DoF robot arm manipulation trajectories from physical-consistent video rollouts.
SAFE-EQA: Semantic-Aware Efficient Exploration for Embodied Question Answering: To tackle the myopic wandering and prohibitive VLM query overhead inherent in greedy frontier-based exploration, SAFE-EQA casts embodied active exploration as an online semantic-aware Traveling Salesman Problem (TSP) over topological Voronoi landmarks, augmented with frontier-triggered adaptive replanning and hierarchical answering, matching or surpassing SOTA accuracy and exploration efficiency while slashing token consumption by 27.8%–45.6%.
Scalable Cross-embodiment Dexterous Grasping via Morphology-Prior Diffusion: By reformulating cross-embodiment dexterous grasping as a 3D assembly problem of rigid links, SOMO couples a shared 3D shape-aware geometric encoder with a classifier-free morphology-prior diffusion model, achieving scalable synthesis across diverse hands and annotation-free zero-shot transfer using only robot URDFs.
Sim, Yet Same: Physics-Aligned Simulator as Zero-Shot Data Scaler in Deformable Worlds: Addressing the vulnerability and geometric-dynamical mismatch of cloth simulation in robotic learning, SIM1 introduces a physics-aligned Real-to-Sim-to-Real data engine that combines sub-millimeter scene digitization, deformation-stable Augmented Vertex Block Descent (AVBD) solving, and diffusion-based motion synthesis, achieving a 15:1 synthetic-to-real data equivalence ratio and a 90% zero-shot real-robot success rate in garment folding.
Stabilizing Real-World Visual Active Tracking with Action-Smooth Test-Time Adaptation: Addressing the severe sim-to-real visual domain shifts and distractor confusion in embodied visual active tracking, this paper proposes VATA, the first test-time adaptation framework for VAT that couples critic value maximization for sample filtering with action smoothness regularization for motion continuity, boosting real-world physical drone tracking success rate from 20% to 80%.
Stand Up and Move: Benchmarking Interactive Spatial Intelligence in WalkerBench: To overcome the limitations of static "Spectator View" benchmarks that ignore active parallax for disambiguation, this paper presents WalkerBench—a global interactive benchmark spanning 161 cities—reveals a fundamental representation misalignment between VLMs' 1D linear context and 3D topology, and introduces Spatial-IDE, a training-free scaffold with explicit topological memory and cognitive decoupling that achieves more than twofold gains on benchmarks and zero-shot transfer on a Unitree humanoid robot.
StructPolicy: Structure-Guided Imitation Learning Robust to Visual Domain Shifts: StructPolicy introduces a domain-invariant Structure Map composed of geometric primitives under physical constraints, utilizing hierarchical structural encoding and affordance-guided pose generation to enable robust and accurate robot manipulation under severe visual domain shifts including lighting, background color, camera viewpoint, and sensor noise.
TaxoGrasp: Taxonomy-Guided Human Grasp Synthesis with Sparse Contact Constraint: Addressing the severe intent ambiguity and contact-semantics decoupling under sparse thumb-contact constraints, TaxoGrasp establishes the Feix grasp taxonomy as an explicit semantic bottleneck across a three-stage "Cognition-Instantiation-Grounding" framework to synthesize intent-consistent, physically plausible, and geometrically executable human grasps.
TDSR-VLA: Transition-aware Denoising Sequence Representations for Vision-Language-Action: TDSR-VLA proposes a decoupled VLA architecture that reuses early sequence representations (SR) and iterative updated sequence representations (USR) produced during the denoising process of a Stable Diffusion 3 Vision Planner, injecting them via layer-wise Prefix-KV conditioning to guide a flow-matching Action Expert for continuous action chunk generation, substantially boosting small-data long-horizon bimanual manipulation.
Towards More Efficient Decoding for Autoregressive Vision-language-action Models: Addressing the severe inference speed bottleneck of autoregressive Vision-Language-Action (VLA) models in robotic control, CEED-VLA introduces consistency distillation on Jacobi trajectories to break token dependencies and pairs it with adaptive early-exit decoding to cut tail redundant iterations, achieving over 4× acceleration across diverse architectures while preserving high manipulation success rates.
Twin-DAgger: Synergizing Digital Twins and Human Corrections for Efficient Robot Manipulation: Twin-DAgger treats sparse real-world human corrections as spatial failure probes and temporal critical priors, enabling difficulty-stratified sampling and intervention-focused trajectory warping within digital twins to match conventional DAgger performance with only 15% of the human intervention cost.
VLA Knows Its Limits: Adaptive Execution Horizons for Robot Policies: This paper discovers that the execution horizon (the number of steps actually executed in each action chunk) of flow-based VLAs exhibits a "rise-then-fall" unimodal performance curve with a unique optimal point. It reveals that the "radial action sinks" at the start and end of the action self-attention encode the model's prediction limits. Based on this, the authors propose AutoHorizon, a training-free and virtually overhead-free test-time method to dynamically estimate the execution horizon for each action chunk.
ZAP: Zero-Shot Assembly Planning with Large Language Models: ZAP first infers a structured "implicit assembly manual" from multi-view part images, then has a vision-language model (VLM) propose the next assembly action and uses digital-twin verification with backtracking search to correct mistakes, reaching an 88.67% planning success rate on 150 ASAP-G assembly tasks — a number that measures the simulation-verified pass rate, not a real-robot success rate at the same scale.