Postdoc @berkeley_ai. Researcher @IBMResearch. PhD @TelAvivUni. Working on Multimodal Robot Foundation Models and Structured Physical AI.
Berkeley, CA
Joined March 2017
- Tweets2.6K
- Following939
- Followers2.1K
- Likes8.9K
Roei Herzig retweeted
Want to see how far a robot policy can go on just a handful of demos?
At #ECCV2026 in Malmö today I'm presenting Robotic Steering: adapt a VLA by tuning only the attention heads that matter. Same performance, fewer parameters, and it holds up on a real robot!
🎤 Spotlight Oral, 9:10, Palissad (Malmö Arena)
🖼️ Poster #344, 10:30–12:30, Exhall
Project Page: chancharikmitra.github.io/ro…
Roei Herzig retweeted
Advice for new PhD students: choose a mentor who is committed to your training, not just your project. The most important product of your PhD - even more important than the science - is you, a newly trained scientist.
Roei Herzig retweeted
What if the best visual reasoning steps are ones humans can’t specify? 🤔
Existing VLM reasoning is often constrained by language, pixels, and human-designed intermediates.
We introduce Latent Implicit Visual Reasoning, where we show that VLMs can discover the best visual reasoning steps by themselves — no bboxes, no intermediate images, no extra supervision.
Presenting this week at CVPR!
(1/n)🧵
Roei Herzig retweeted
Check out our new work led by @Dantong_Niu: T-Rex = EgoScale + tactile sensing.
My view: tactile, force, and contact will be critical interfaces for human-to-robot transfer. We still lack good human tactile sensorization, but one step at a time.
Excited to share T-Rex: Tactile-Reactive Dexterous Manipulation 🦖🤖
Touch is fundamental to human dexterity, yet most Vision-Language-Action (VLA) models either ignore tactile feedback or lack the ability to react to high-frequency contact signals.
In this work, we tackle both the data and architectural challenges of tactile-reactive dexterous manipulation.
🦖 A 100-hour tactile-synchronized dexterous manipulation dataset with 7,700+ trajectories, 22 motor primitives, and 200+ everyday objects.
🦖 A tactile-reactive MoT architecture with spatial-temporal tactile encoding and asynchronous high-frequency tactile refinement.
🦖 A scalable training recipe combining 22,889 hours of human egocentric pretraining with tactile-grounded robot mid-training.
Across 12 real-world contact-rich manipulation tasks, T-Rex achieves over 30% higher average success rate than the strongest baseline.
We are fully open-sourcing the dataset, models, teleoperation stack, training code, and inference pipeline.
🌐 Project: tactile-rex.github.io/
📄 Paper: arxiv.org/abs/2606.17055
💻 Code: github.com/ZhuoyangLiu2005/T…
🤗 Dataset: huggingface.co/datasets/zeka…
🧵 Thread ↓
Roei Herzig retweeted
Researchers from @nvidia , the @UofCalifornia Berkeley and @Stanford used @ManusMeta gloves to capture precise finger motion for the T‑Rex framework.
The framework unifies vision, language and tactile sensing for real‑time contact‑aware robot behaviour.
The foundation is a 100‑hour tactile‑synchronised teleoperation dataset spanning more than 200 everyday objects and 22 motor primitives.
Learn more about T-Rex: tactile-rex.github.io/
More about MANUS™:
humanoidroboticstechnology.c…
#humanoids #humanoidrobotics #humanoidrobot #humanoid
Roei Herzig retweeted
Excited to share T-Rex: Tactile-Reactive Dexterous Manipulation 🦖🤖
Touch is fundamental to human dexterity, yet most Vision-Language-Action (VLA) models either ignore tactile feedback or lack the ability to react to high-frequency contact signals.
In this work, we tackle both the data and architectural challenges of tactile-reactive dexterous manipulation.
🦖 A 100-hour tactile-synchronized dexterous manipulation dataset with 7,700+ trajectories, 22 motor primitives, and 200+ everyday objects.
🦖 A tactile-reactive MoT architecture with spatial-temporal tactile encoding and asynchronous high-frequency tactile refinement.
🦖 A scalable training recipe combining 22,889 hours of human egocentric pretraining with tactile-grounded robot mid-training.
Across 12 real-world contact-rich manipulation tasks, T-Rex achieves over 30% higher average success rate than the strongest baseline.
We are fully open-sourcing the dataset, models, teleoperation stack, training code, and inference pipeline.
🌐 Project: tactile-rex.github.io/
📄 Paper: arxiv.org/abs/2606.17055
💻 Code: github.com/ZhuoyangLiu2005/T…
🤗 Dataset: huggingface.co/datasets/zeka…
🧵 Thread ↓
Roei Herzig retweeted
When a robot learns to react to touch, where does the training data come from?
Researchers from @UCBerkeley, @nvidia, and @Stanford introduce T-Rex, a framework that unifies vision, language, and tactile sensing so robots can respond to physical contact in real time rather than relying on vision alone.
On contact-rich tasks like inserting a card, turning a key, and handling deformable objects, it outperforms the strongest baseline by more than 30% across 12 real-world tasks.
The foundation is a 100-hour tactile-synchronized teleoperation dataset spanning 200+ everyday objects and 22 motor primitives. During data collection, researchers wore @ManusMeta gloves to capture precise finger motion, which was then retargeted onto @SharpaRobotics Wave dexterous hands for bimanual teleoperation.
Learn more: tactile-rex.github.io/
Roei Herzig retweeted
The videos from the “Frontiers of Embodied AI” meetup at ETHZ from a few weeks back are now available.
Speakers: Jitendra Malik, Vladlen Koltun, Yann LeCun, and Shuran Song
Hosted by Marc Pollefeys
YouTube playlist: youtube.com/playlist?list=PL… – at Zurich, Switzerland
Roei Herzig retweeted
Full video - youtube.com/live/cFwHAMf0kEc…
Roei Herzig retweeted
#ICRA2026 will feel like my grad school days, where I know few people 😬 A goal is to meet new people and extend my network.
Don’t be shy, say hi!
Roei Herzig retweeted
Relative to native-born Americans, several countries' immigrants to America have produced vastly greater rates of unicorn founders.
In particular, Israel: with shockingly few people in the U.S., Israel is the #2 origin *in total* for foreign Unicorn founders.
Roei Herzig retweeted
When people share a space, their movements become intertwined. Embodied agents need to understand these social dynamics to interact effectively.
Introducing MAGNet 🧲, a unified autoregressive diffusion forcing model for multi-agent motion generation that captures these interactions.
MAGNet is flexible: predict the future, fill in missing motion, or have people react to each other, all while naturally scaling to N>2 people and generating ultra-long motion sequences.
Roei Herzig retweeted
📢 Deadline Extension for MMFM Workshop @ #CVPR2026!
We are extending the submission deadline to **March 14, 2026 (AoE)**. For updated details on submission timelines and guidelines, please refer to the workshop website and OpenReview page below. We’re excited to see your work!
The 5th edition of the MMFM Workshop is coming to @CVPR 2026!
"What is Next in Multimodal Foundation Models?" exploring the frontiers of vision, language, and beyond.
June 2026 | Denver, CO
Details in thread 👇
Roei Herzig retweeted
TLDR: Your SSL model behaves like this:
1. Directional neural collapse: within-class variance shrinks along class-separating directions, even if it stays large elsewhere.
2. Task orthogonalization: different downstream tasks tend to use nearly orthogonal decision directions.
Roei Herzig retweeted
Our deadline is only one week away! Don't forget to submit!
The 5th edition of the MMFM Workshop is coming to @CVPR 2026!
"What is Next in Multimodal Foundation Models?" exploring the frontiers of vision, language, and beyond.
June 2026 | Denver, CO
Details in thread 👇
Roei Herzig retweeted
📄 Paper: arxiv.org/pdf/2511.22697
🌐 Project page: chancharikmitra.github.io/ro…
Huge thanks to our amazing team across @CMU_Robotics , @berkeley_ai , and @USCViterbi 🙏!
Yusen Luo @yusen_2001 , Raj Saravanan, Dantong Niu @Dantong_Niu , Anirudh Pai, Jesse Thomason @_jessethomason_ , Trevor Darrell @trevordarrell, Abrar Anwar @_abraranwar , Deva Ramanan @RamananDeva , Roei Herzig @roeiherzig
