BeingBeyond/Being-H

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1126 stars · 61 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Being-H is a family of human-centric embodied foundation models, including VLA models (Being-H0.5, Being-H0) and latent world-action models (Being-H0.7, Being-H0.8) pretrained on large-scale egocentric human manipulation video. The repository provides training, inference, post-training, and deployment code plus pretrained checkpoints on Hugging Face for dexterous robot manipulation.

Use cases

• train a vision-language-action model for dexterous robot hands
• pretrain a world-action model on egocentric human videos
• post-train a VLA model on a small amount of real robot data
• achieve cross-embodiment generalization for manipulation tasks
• generate tactile-aware contact predictions for grasping and insertion
• deploy pretrained manipulation models on humanoid robots or dexterous hands

When to choose

• you need a human-video-pretrained VLA or world-action model for dexterous manipulation
• you want cross-embodiment transfer with minimal real-robot fine-tuning
• you are researching tactile or latent world models for embodied AI

When to avoid

• you need a simple off-the-shelf robot control stack without model training
• your hardware lacks GPUs or the robotics setup the models target
• you need a mature production system rather than a research codebase

Facets

library · maturity active

machine-learning deep-learning llm-training simulation robotics machine-learning artificial-intelligence computer-vision python embodied-ai vla world-model dexterous-manipulation foundation-model egocentric-video tactile humanoid-robotics cross-embodiment linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="BeingBeyond/Being-H")