Research Scientist at @GoogleDeepMind Gemini Robotics 🦾🤖| @UMRobotics Alumni | Interned at @AIatMeta @NVIDIARobotics @amazondrives
Joined January 2022
- Tweets82
- Following204
- Followers576
- Likes494
Pinned Tweet
Dexterous hands vary widely—so do tactile modalities. 🖐️🌈
Our vision on tactile human-to-robot transfer:
🔓 Not tied to specific hardware
♻️ Reuse human tactile demos across embodiments
Presenting TactAlign, a cross-sensor tactile alignment for cross-embodiment policy transfer.
Super cool to see Argon’s real2sim2real pipeline iteratively refine a dynamic whole-body policy! 🚀 Exciting to see how Argon can accelerate robot skill acquisition and unlock new capabilities!
Gemini 4 Argon taught a robot to play cornhole just by watching a video.
It mapped the physics, coded its own simulation, and taught @Apptronik's Apollo humanoid to sink 7 throws in a row.
I believe human dexterity is not just about precise motor control, but even more about mastering compliance :) I’ve been thinking a lot about how we can achieve this in high-DoF systems, so I’m super excited to read this paper! Congrats! @HaozhiQ @Ying_yyyyyyyy
It's interesting that I have spent quite some time working on sim-to-real for screwdriving, often using it as an example of why finger coordination is too hard to teleoperate and why we need simulation.
But is that really the case?
We explore whether carefully designed low-level control, especially with compliance, can actually make these tasks teleoperable. Checkout the original thread ⬇️
And the @SharpaRobotics glove has been amazing to work with! It played a key role in making this project possible!
Youngsun Wi ✈️ CoRL 2026 retweeted
We asked #Gemini 4 Argon to explain how it figured out how to play cornhole with @Apptronik's Apollo - here it is in its own words 🤯. This is transformative for Physical AI 🦾.
Try it yourself at : lnkd.in/g9WyK_CC
Learn more: lnkd.in/gX9NeEjx
@GoogleDeepMind #Robotics
Youngsun Wi ✈️ CoRL 2026 retweeted
🧤Excited to share that the OSMO tactile glove is now available as a DIY kit!
OSMO is our open-source tactile glove for human-to-robot skill transfer, capturing 3-axis tactile signals from human demonstrations and enabling the same sensing hardware to be used on robot hands.
Huge thanks to Jessica Yin (not on X) for driving this effort and making the kit happen, and to @WowRobo for helping make OSMO more accessible to the community!
(Neither I nor any of the authors receive money or profit from kit sales. Our goal is to support reproducibility and help advance the tactile research community. If you find our work helpful, we just ask that you cite our work 😊 The hardware designs, firmware, and assembly instructions remain publicly available:)
🔧 DIY kit: shop.wowrobo.com/products/os…
📖 Project: jessicayin.com/osmo_tactile_…
w. Jessica Yin, @WiYoungsun, Sayantan Kundu, Mike Lambeta, William Yang, @ChanghaoWang78, @TingfanW1208, @JitendraMalikCV, and @t_hellebrekers!
Youngsun Wi ✈️ CoRL 2026 retweeted
Transformer Transformer will be presented at #CoRL2026 🚀 See y'all in Austin!
Website: transformer-transformer.gith…
Video: youtube.com/watch?v=TTyjvPVF…
Paper: transformer-transformer.gith…
Youngsun Wi ✈️ CoRL 2026 retweeted
One brain. For any robot. 🤖
We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more.
Excited to present TactAlign at RSS today!
How can robots learn from highly sensitized human demo (touch + proprio) despite “differences in tactile sensors”?
Humans are an incredible source of dexterous data. Let’s talk about how we can make the most of it!
I'm at RSS this year in Sydney!
I've been thinking a lot these days about tactile sim-to-real, exoskeletons, forceful manipulation, and tactile retargeting to leverage human data. Dms open and always open to talk about these topics. I believe the next frontier of robot learning is forceful.
We're also presenting TactAlign (lead author @WiYoungsun) today in manipulation session 1 (July 13, 2:30pm - 5:15pm).
Youngsun Wi ✈️ CoRL 2026 retweeted
I'm at RSS this year in Sydney!
I've been thinking a lot these days about tactile sim-to-real, exoskeletons, forceful manipulation, and tactile retargeting to leverage human data. Dms open and always open to talk about these topics. I believe the next frontier of robot learning is forceful.
We're also presenting TactAlign (lead author @WiYoungsun) today in manipulation session 1 (July 13, 2:30pm - 5:15pm).
Youngsun Wi ✈️ CoRL 2026 retweeted
Do VLMs actually understand 3D space 🌎?
Or are they exploiting shortcuts hidden in natural images?
🚀 Excited to share our new work:
Why Far Looks Up: Probing Spatial Representation in Vision-Language Models
@NVIDIAAI × @SeoulNatlUni × @OhioStateCSE
🧵👇
Youngsun Wi ✈️ CoRL 2026 retweeted
Human data collection roughly 4x faster than robot teleoperation:
A method that aligns human glove-based tactile signals with robot sensors via self-supervised encoders and rectified flow using motion-derived pseudo-pairs, avoiding any need for paired data or identical hardware.
The approach enables human-to-robot policy transfer with only 5-10 minutes of human demos per object, co-training policies for contact-rich tasks like pivoting, insertion, and lid-closing to achieve major gains in generalization to unseen objects.
Key result includes zero-shot transfer to dexterous tasks such as lightbulb screwing by reusing tactile alignment learned from prior tasks, with the mentioned human data collection roughly 4x faster than robot teleoperation.
Thanks for sharing, Youngsun Wi (@WiYoungsun)
📌 Official project website (full details, videos, results, and method overview):
yswi.github.io/tactalign/
——-
Weekly robotics and AI insights.
Subscribe free: 22astronauts.com
TactAlign was accepted to RSS 2026! Huge thanks to the reviewers for their thoughtful feedback. See you on the other side of the world 🤓🙌
Youngsun Wi ✈️ CoRL 2026 retweeted
🎊Excited to announce the ERA challenge for #CVPR2026 !
Please checkout the details below 👇
We have $6K in cash prizes, thanks to our sponsors: @NVIDIARobotics, @BitRobotNetwork, and @cortexairobot
Excited to share Embodied Reasoning in Action at @CVPR 2026 — a workshop and open challenge on robotic manipulation.
We’re launching two challenges:
⚡ RoboSpatial — point, fit, and place from RGB-D home scenes
⚡ PointArena — language-guided pixel pointing for VLM evaluation
💰 Up to $6K in cash prizes
📍 Denver, CO
📅 June 3–4, 2026
🗓️ Challenge deadline: May 23, 2026
Huge thanks to our sponsors: @NVIDIARobotics , @BitRobotNetwork , and @cortexairobot
Learn more: embodied-reasoning.github.io
Youngsun Wi ✈️ CoRL 2026 retweeted
We’re rolling out an upgrade designed to help robots reason about the physical world. 🤖
Gemini Robotics-ER 1.6 has significantly better visual and spatial understanding in order to plan and complete more useful tasks. Here’s why this is important 🧵
Very cool to see full-hand tactile sensing @binghao_huang 👀
In TactAlign, we showed human-to-robot tactile transfer with fingertip sensing.
Now that we have FlexiTac on the robot, this feels like a great time to extend TactAlign to “full-palm” human-to-robot tactile transfer!!
🤲Tactile sensing is powerful for robot manipulation, but hardware is still difficult to access, reproduce, and scale.
🎯That’s why we built FlexiTac: an open-source, low-cost, and scalable tactile sensing solution designed for real robotic systems.
• Project page: flexitac.github.io/
We hope FlexiTac can help democratize tactile sensing for robotics research. (1/n)
Youngsun Wi ✈️ CoRL 2026 retweeted
🧠 Excited to share our vision for embodied systems: generalist models orchestrating specialist perception tools for spatial reasoning.
🛠️ Even robots themselves are treated as tools for manipulation.
🚀 Great work led by @ChenSiyich and the team at @nvidia behind RoboSpatial!
🎉 Accepted to #CVPR2026
🔎 VLMs fall short on complex spatial reasoning.
They struggle with:
• Precise geometric perception
• Multi-step reasoning grounded in 3D
• Adapting perception dynamically to task and context
🚀 We propose a solution: visual tool-augmented spatial reasoning —
bridging perception and multi-step reasoning through diverse, error-aware, adaptive vision tool use.
And we go one step further:
🤖 enabling robot control by treating robots themselves as tools.
Our framework is powered by:
⚡ Double Interactive RL (DIRL), a new training framework combining demonstrations + real exploration
🛠 Real interaction with specialized computer vision models during RL
🤖 Toolshed, a scalable, asynchronous system for multimodal execution of vision tools and robots-as-tools
🔗 Project: spacetools.github.io/
</> Code: github.com/spacetools/SpaceT…
Toolshed is released with frontier-model demos.
Full training & evaluation release coming soon.
Done during my internship @NVIDIA — big thanks to the amazing collaborators! 🙌
#CVPR2026 #EmbodiedAI #ComputerVision #Robotics #ReinforcementLearning #MultimodalAI
Youngsun Wi ✈️ CoRL 2026 retweeted
Most world models for robot manipulation learn physics from pixels. But pixels don’t see it all.
Can we ground these models in the "feeling" of contact to disambiguate visually identical states?
Visuo-Tactile World Models (VT-WM): robot imagination in a shared space👇
Youngsun Wi ✈️ CoRL 2026 retweeted
🚀 Freshly accepted to CVPR 2026
What if we could train computer-using agents just by watching YouTube?
We present Watch & Learn (W&L) -- a inverse-dynamics framework that turns internet videos of humans using computers into learnable UI trajectories at scale.
Thread 👇
Check out the website ( yswi.github.io/tactalign/) for more details!
Thanks so much for wonderful collaborators and mentors: Jessica Yin, @elviskxiang, @akashshrm02 , @JitendraMalikCV, @mukadammh, @NimaFazeli7, @t_hellebrekers
Also, huge thanks to everyone who has provided valuable input and support throughout: @luke_ch_song, Fan Yang, @MarkVanderMerwe, James Lorenz, @TingfanW1208, Francois Hogan, Taosha Fan, Sayantan Kundu, Mike Lambeta, Wonik Robotics.