[NeurIPS 2024] PyTorch code for the paper "Making Offline RL Online: Collaborative World Models for Offline Visual Reinforcement Learning"
Experiments on Model-Agnostic Meta-Learning on Few-Shot Image Classification and Meta-RL (Meta-World)
A simple, efficient, and extensible evaluation framework for robot policies across diverse benchmarks, supporting both local and remote execution
Repository of the paper: Projected Task-Specific Layers for Multi-Task Reinforcement Learning
A lightweight Vision-Language-Action (VLA) baseline for MetaWorld robot-arm tasks using a pretrained CLIP-ViT vision transformer(openai/clip-vit-base-patch32), a small text transformer, and robot-state fusion
Official implementation of VGGT-DP: Generalizable Robot Control via Vision Foundation Models (arXiv:2509.18778). A diffusion policy that uses the frozen VGGT vision foundation model as a geometry-aware visual encoder for robot manipulation, evaluated on MetaWorld.
ICLR 2026 paper on state-dependent reward shaping with multi-view videos and ViCLIP for reinforcement learning.
An implementation of Metaworld with Garage
A world model as a judge: early, auditable verdicts on partial robot-manipulation rollouts
Access Oxtia on ChatGPT.x
Hash-pinned full-corpus audit and guarded repair preview for the LeRobot MetaWorld MT50 task-label collision
Adaptive replan-gate for frozen flow-matching VLAs on MetaWorld
To associate your repository with the metaworld topic, visit your repo's landing page and select "manage topics."