How can we effectively leverage human videos for robot learning by bridging the inherent embodiment gap? We introduce UniSkill, a universal skill representation, a scalable method for learning cross-embodiment skill representations from large-scale in-the-wild video data. 1/n

When @anyazorin and @irmakkguzey open-sourced the RUKA Hand (a low-cost robotic hand) earlier this year, people kept asking us how to get one. Open hardware isn’t as easy to share as code. So we’re releasing an off-the-shelf RUKA, in collaboration with @WowRobo and @zhazhali01.

NEO The Home Robot Order Today

MLLMs show strong video QA ability, but how can we extend this to the 𝐩𝐢𝐱𝐞𝐥-𝐥𝐞𝐯𝐞𝐥 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐚𝐧𝐲 𝐭𝐫𝐚𝐢𝐧𝐢𝐧𝐠? We introduce Decomposed Attention Fusion (DecAF) in MLLMs for Training-Free Video Reasoning Segmentation.

Introducing Figure 03 https://t.co/cgJtQkYb3p

Excited for CoRL! Heading from NYC back to Korea 🇰🇷 I'll present UniSkill at the H2R workshop (Sat) and Poster Session (Mon) 🤖 If you're into Learning from Human Video or Latent Action Modeling, I'd be happy to chat!

Unitree G1 has mastered more quirky skills 🤩 Unitree G1 has learned the "Anti-Gravity" mode: stability is greatly improved under any action sequence, and even if it falls, it can quickly get back up.

Coming from a computer vision background and now in sequence modeling, I’m often struck by how disconnected LLMs and vision feel. Our work, AUSM, treats video as language -- and it reveals a few blind spots we’ve overlooked.

We connect the autoregressive pipeline of LLMs with streaming video perception. Introducing AUSM: Autoregressive Universal Video Segmentation Model. A step toward unified, scalable video perception — inspired by how LLMs unified NLP. 📝

Learn robot skills from human YouTube videos, without… labels, poses, or human-robot alignment: 📌 Bookmarked by researchers pushing the limits of cross-embodiment learning. UniSkill is a new framework for learning embodiment-agnostic skill representations from large-scale,

Happy to announce our paper, UniSkill, is accepted to CoRL 2025! 🤖 Learning from human videos is the future of robot learning, but the cross-embodiment gap has been a major barrier. We introduce a simple yet powerful way to bridge this gap. Looking forward to see in Korea!🇰🇷

Tokenization has been the final barrier to truly end-to-end language models. We developed the H-Net: a hierarchical network that replaces tokenization with a dynamic chunking process directly inside the model, automatically discovering and operating over meaningful units of data

🎞️ 𝐃𝐨𝐮𝐛𝐥𝐞 𝐭𝐡𝐞 𝐒𝐩𝐞𝐞𝐝, 𝐙𝐞𝐫𝐨 𝐓𝐫𝐚𝐢𝐧𝐢𝐧𝐠: 𝐓𝐡𝐞 𝐅𝐫𝐞𝐞 𝐋𝐮𝐧𝐜𝐡 𝐟𝐨𝐫 𝐕𝐢𝐝𝐞𝐨 𝐋𝐋𝐌𝐬! ⚡️ 🚨I am excited to share that our paper is accepted #ICCV2025 @ICCVConference ArXiv paper: https://t.co/8dnqmHoAsm Project page:

Generalization needs data. But data collection is hard for precise tasks like plugging USBs, swiping cards, inserting plugs, and keying locks. Introducing robust, precise VisuoTactile Local (ViTaL) policies: >90% success rates from just 30 demos and 45 min of real-world RL.🧶⬇️

Tactile sensing is gaining traction, but slowly. Why? Because integration remains difficult. But what if adding touch sensors to your robot was as easy as hitting “print”? Introducing eFlesh: a 3D-printable, customizable tactile sensor. Shape it. Size it. Print it. 🧶👇

It was nice engaging with the CV community on ways to stand out in the crowd. My answer was simple: work on robotics. There are so many unanswered problems and open pastures for research if you are a new researcher. Below are 6 problems I focussed on in my talk.

Live demo-ing RUMs at @CVPR this afternoon next to the expo sessions – stop by with something small and let’s see if the robot can pick it up zero shot! #CVPR2025

We just open-sourced EgoZero! It includes the full preprocessing to turn long-form recordings into individual demonstrations as 3D states + actions. We engineered this for scalability to big datasets (streaming, parallel workers, CPU/GPU utilization)

Everyday human data is robotics’ answer to internet-scale tokens. But how can robots learn to feel—just from videos?📹 Introducing FeelTheForce (FTF): force-sensitive manipulation policies learned from natural human interactions🖐️🤖 👉 https://t.co/CZcG87xYn5 1/n

Teaching robots to learn only from RGB human videos is hard! In Feel The Force (FTF), we teach robots to mimic the tactile feedback humans experience when handling objects. This allows for delicate, touch-sensitive tasks—like picking up a raw egg without breaking it. 🧵👇

UniSkill is accepted at @CVPR Agents in Interactions, from Humans to Robots workshop. I'll be attending CVPR—would like to connect and chat with folks in the robotics. Feel free to ping me!