Juncheng Mu

Can robots learn a bimanual dexterous skill from just one video? DexImit makes it possible 🏄‍♂️ — automatically converting human videos into physically plausible training data for bimanual manipulation, either from text2video models or real-world recordings.

1/ We just released π0.7 — a steerable generalist robot model with emergent capabilities. I want to share a bit of the backstory, because π0.7 taught me something surprising about where robot learning is heading. A thread on bittersweet lessons 🧵

Introducing GEN-1. Our latest milestone in scaling robot learning. We believe it to be the first general-purpose AI model to master simple physical tasks. 99% success rates, 3x faster speeds, adapts in real time to unexpected scenarios, w/ only 1 hour of robot data. More🧵👇

@ju_yuanchen Thanks Yuanchen — really appreciate it! Glad you found it interesting 🥳

🤩A scalable framework that converts human videos into physically grounded bimanual manipulation data — and transfers to real robots. Really exciting work. Congrats! Juncheng！

@andrew_n_carr Thanks — really appreciate it !!!

@Juncheng_Mu Wow! Really stunning work

Can robots learn a bimanual dexterous skill from just one video? DexImit makes it possible 🏄‍♂️ — automatically converting human videos into physically plausible training data for bimanual manipulation, either from text2video models or real-world recordings.

Huge thanks to my amazing collaborators! 🔥 @SizheYang111 @still_wtm @EvenEveno Big shoutout to our best advisors: @Boyiliee @HarryXu12 @pangjiangmiao 🙌

Learn more and explore our project 👇 📜 Paper: arxiv.org/abs/2602.10105 🌐 Project Page: mujc2021.github.io/deximit/

📊 Result We evaluate DexImit along two dimensions: input video quality and task difficulty. DexImit can generate diverse short-horizon data using only a text2video model. For long-horizon tasks such as beverage making and apple cutting, higher-quality videos are necessary.