Meta-World

Meta-World is a hierarchical world model that integrates semantic planning and physical control for humanoid robot loco-manipulation, used to evaluate skill transfer and composition in response to high-level instructions.

Papers using Meta-World (41)

Demonstration-free Robotic Control Via LLM Agents2026Probeflow: Training-free Adaptive Flow Matching For Vision-language-action Models2026Metaworld: Skill Transfer And Composition In A Hierarchical World Model For Grounding High-level Instructions2026Oflow: Injecting Object-aware Temporal Flow Matching For Robust Robotic Manipulation2026Pocketdp3: Efficient Pocket-scale 3D Visuomotor Policy2026MARVL: Multi-stage Guidance For Robotic Manipulation Via Vision-language Models2026Hyper-DP3: Frequency-Aware Right-Sizing of 3D Diffusion Policies for Visuomotor Control2026Learning Bilevel Policies over Symbolic World Models for Long-Horizon Planning2026Information Filtering via Variational Regularization for Robot Manipulation2026Merging And Disentangling Views In Visual Reinforcement Learning For Robotic Manipulation2025Data-efficient Multitask Dagger2025Coupled Distributional Random Expert Distillation For World Model Online Imitation Learning2025Timerewarder: Learning Dense Reward From Passive Videos Via Frame-wise Temporal Distance2025Self-improving Loops For Visual Robotic Planning2025Pixel Motion Diffusion Is What We Need For Robot Control20253D Flow Diffusion Policy: Visuomotor Policy Learning Via Generating Flow In 3D Space2025ISS Policy : Scalable Diffusion Policy With Implicit Scene Supervision2025Masked Generative Policy For Robotic Control2025Merging and Disentangling Views in Visual Reinforcement Learning for Robotic Manipulation2025Time Reversal Symmetry for Efficient Robotic Manipulations in Deep Reinforcement Learning2025Self-Improving Loops for Visual Robotic Planning20253D Flow Diffusion Policy: Visuomotor Policy Learning via Generating Flow in 3D Space2025VGGT-DP: Generalizable Robot Control via Vision Foundation Models2025Trajectory Conditioned Cross-embodiment Skill Transfer2025ISS Policy : Scalable Diffusion Policy with Implicit Scene Supervision2025Demonstration-Free Robotic Control via LLM Agents2026PocketDP3: Efficient Pocket-Scale 3D Visuomotor Policy2026KAN We Flow? Advancing Robotic Manipulation with 3D Flow Matching via KAN & RWKV2026PRISM: Performer RS-IMLE for Single-pass Multisensory Imitation Learning2026MARVL: Multi-Stage Guidance for Robotic Manipulation via Vision-Language Models2026Primary-Fine Decoupling for Action Generation in Robotic Imitation2026Ada3Drift: Adaptive Training-Time Drifting for One-Step 3D Visuomotor Robotic Manipulation2026ProbeFlow: Training-Free Adaptive Flow Matching for Vision-Language-Action Models2026Multi-View Video Diffusion Policy: A 3D Spatio-Temporal-Aware Video Action Model2026OFlow: Injecting Object-Aware Temporal Flow Matching for Robust Robotic Manipulation2026FRMD: Fast Robot Motion Diffusion with Consistency-Distilled Movement Primitives for Smooth Action Generation2025EC-Flow: Enabling Versatile Robotic Manipulation from Action-Unlabeled Videos via Embodiment-Centric Flow2025MP1: MeanFlow Tames Policy Learning in 1-step for Robotic Manipulation2025Spatial Policy: Guiding Visuomotor Robotic Manipulation with Spatial-Aware Modeling and Reasoning2025Growing with Your Embodied Agent: A Human-in-the-Loop Lifelong Code Generation Framework for Long-Horizon Manipulation Skills2025Pixel Motion Diffusion is What We Need for Robot Control2025