HumanCLAW

Can a VLM act through a body? Open-source benchmark for action intelligence, decoupled from motor control.

by @KuvviusAdded Aug 20, 2026

open source ai robotics research

Jiawei Gu

@Kuvvius

·Follow

🔥 HumanCLAW is now fully open-sourced! github.com/Human-CLAW/Hum…

Jiawei Gu

@Kuvvius

🦾HumanCLAW: Can Vision-Language Models Act Through a Body? A VLM can spot the sofa in a second. But can it get a body there and sit down? We call it Action Intelligence, and our work decouples it from motor control so it can finally be measured. Today’s best models turn out to

4:41 AM · Aug 18, 2026

100

Read 1 reply

Visit project View post on X

Related builds

DexGPT

One video of a human hand becomes a real2sim pipeline. Tracking, IK to a 44-DoF hand, grasp refine. First-party, OSS.

xiao hu

@huxiao93612565

·Follow

Video in → physics out. 🤖 I gave GPT-6 Astra one video of a human hand. It wrote this entire real2sim pipeline itself — hand tracking, IK retargeting to a 44-DOF dexterous hand, grasp refinement. Zero code from me. Open-sourcing everything 👇 github.com/Hu-xiao-max/de…

Watch on X

10:31 PM · Sep 9, 2026

141

Read 5 replies

ai robotics open source

Visit project

Sep 16, 2026

EgoWAM

World action model that lets robot policy scale with in-the-wild human video. CoRL 2026. Code and data are public.

Baoyu Li

@BaoyuLi6

·Follow

EgoWAM has been accepted to #CoRL2026 @corl_conf 🥳🥳 See you all in Austin!!!! We’ve also open-sourced all the data and code: 📊 Data: partners.mecka.ai/egoverse?task=… 💻 Code: github.com/GaTech-RL2/Ego… If you run into any problems, please submit an issue on GitHub. We’ll do our best to Show more

Baoyu Li

@BaoyuLi6

Egocentric human data is abundant, but human motion is not always positive supervision for robot policy due to embodiment gaps. Naive BC co-training can HURT performance ☹️. 🌟Our key finding in **EgoWAM**: the state-prediction branch of a World Action Model effectively bridges

Watch on X

6:54 PM · Sep 7, 2026

Read 2 replies

robotics ai open source

Visit project

Sep 16, 2026

Amplio

DeepMind's harness for autonomous research runs that last days to weeks. Used internally on Simply, now public.

Chen Liang

@crazydonkey200

·Follow

Glad that others are finding Amplio helpful. It is our main harness for autonomous long research runs spanning days to weeks. It is heavily used by our team for automated AI research / RSI on top of the Simply codebase. Open sourced at: github.com/google-deepmin…

Wenhao Chai

@wenhaocha1

Harness matters. We test Gemini 3.1 Pro with amplio, a harness for robust and long-horizon runs and super fit for auto-research. We then get much better results than with Gemini CLI. Harness: github.com/google-deepmin… FrontierCS is an open-ended benchmark for Computer Science

9:47 PM · Aug 28, 2026

Read 1 reply

open source ai agents research

Visit project

Aug 29, 2026

Cyclo Control pick-and-place

ROBOTIS OMY arm: Gemini reasoning, GraspNet 6-DoF poses, collision-free pick-and-place on ROS 2.

Yoonseok Pyo

@passionvirus

·Follow

Smart pick-and-place with OMY, Gemini, and GraspNet! We integrated VLM reasoning, GraspNet 6-DoF pose synthesis, and Cyclo Control for collision-free manipulation on ROS 2. 🎬 Video: youtu.be/OxMBVGkFJHM 🤖 Code: github.com/ROBOTIS-GIT/cy… 📝 Docs: docs.robotis.com/docs/systems/o…

Watch on X

12:57 PM · Aug 28, 2026