cross-embodiment

HEX is a whole-body vision-language-action framework for full-sized humanoid robots.

LAP: Language-Action Pre-Training Enables Zero-Shot Cross Embodiment Transfer

LocoFormer - Generalist Locomotion via Long-Context Adaptation

Official implementation of the ICML 2026 paper "DiLA: Disentangled Latent Action World Models".

Awesome robot data engines for VLA: collection, synthesis, augmentation, curation, preprocessing, and benchmarks.

Repository hosting the official code of the paper "PCHands: PCA-based Hand Pose Retargeting on Manipulators with N-DoF"

🤖 Explore LocoFormer, a Transformer-XL model that enhances robot locomotion through long-context learning and real-world adaptability.

Curated papers, datasets, systems, and benchmarks for robot data engines across robot-centric, UMI, human/egocentric, and simulation data.

Paper/Code list of cross-emboided.

Answers whether several robot demonstration datasets can be mixed by inspecting their manifests alone, scoring mixability from action space, control frequency, gripper convention, and metadata gaps.

X-Embodiment Language-Grounded Manipulation Benchmark: measuring language-conditioned policy transfer between a Panda arm and a Unitree G1 humanoid in ManiSkill3