Models

Research

UniAct studies universal action representations for embodied foundation models.

Research

Natural Locomotion, Jumping and Recovery of Quadruped Robot A1 with AMP

Research

X-VLA studies soft-prompted transformers for cross-embodiment vision-language-action learning.

Research

HiMoE-VLA studies hierarchical mixture-of-experts policies for vision-language-action learning.

Research

DreamDojo studies robot world models learned from human video data.

Research

Official Implementation for the paper "SR-AIF: Solving Sparse-Reward Robotic Tasks from Pixels with Active Inference and World Models"

Research

SimpleVLA-RL provides reinforcement-learning training resources for VLA policies.

Research

VLAC: A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learning

Research

VLA-RL provides reinforcement-learning code for adapting VLA models to robotic manipulation.

Open source

Building General-Purpose Robots Based on Embodied Foundation Model

Research

LingBot-VLA provides a vision-language-action foundation-model project.

Open source

Codebase for DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation

Research

FOCA studies future-oriented conditioning for data-efficient VLA adaptation.

Open source

A Minimalist Retargeting-Guided Reinforcement Learning Recipe for Dexterous Manipulation

Open source

Visual Imitation Enables Contextual Humanoid Control. CoRL 2025, Best Student Paper Award.

Research

A robot policy project using world-model representations.

Research

Uni-NaVid: A Video-based Vision-Language-Action Model for Unifying Embodied Navigation Tasks.

Research

world modeling challenge for humanoid robots

Research

Language-guided 3D value-map planning for robot manipulation.

Research

Being-H is BeingBeyond's family of human-centric embodied foundation models.

Research

From Vision-Language-Action Models to a Real-World Robot Learning Stack

Research

VLA-0 provides research code for building vision-language-action models from vision-language models.

Research

A unified framework for training, fine-tuning, and evaluating World Action Models

Research

ICCV 2025 | TesserAct: Learning 4D Embodied World Models