Ron Vainshtein (@ronvain) profile picture

Computer Vision Researcher.

Joined April 2022
Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)Photo by Ron Vainshtein (@ronvain)
Ron Vainshtein retweeted
Animation ๐Ÿค Robotics ProtoMotions GTC 2026 release โ€” bridging the gap between digital humans and real humanoid robots. Train in simulation. Deploy on hardware. One framework, one codebase. nvlabs.github.io/ProtoMotionโ€ฆ
6
31
1
132
9,915
Ron Vainshtein retweeted
Just arrived at San Diego for #Neurips2025 โœˆ๏ธ If you're interested in - Tactile sensing ๐Ÿซฐ - Uncertain decision making โ†”๏ธ - Medical robotics ๐Ÿฉบ Reach out or come see if you process touch better than AI at our poster on Fri 16:30-19:30 #2206. @NeurIPSConf
Can we use tactile sensors for soft object imaging?๐Ÿค” YES! Toward Artificial Palpation: Representation Learning of Touch on Soft Bodies. @shafer_eli , @talt80, @Efrat_Shimron, @AvivTamar1. Paper, videos, data: zoharri.github.io/artificialโ€ฆ Check us out at @NeurIPSConf #neurips2025!
3
11
684
Ron Vainshtein retweeted
1. Train a behavior foundation model with a rich token-based representation (eg, MaskedMimic). 2. Fine tune a state-based tokenizer using RL, leveraging the BFM for meaningful gradients. => Converge 6x faster, 50x less params, dramatically improved robustness.
1
9
2
57
6,809
Ron Vainshtein retweeted
Behavior models are great โ€” but getting them to do *exactly* what you want takes *tons* of prompt hacking (just like vision-language models) ๐Ÿ˜ช No more, with our new work ๐Ÿ“ฐ - Task Tokens: A Flexible Approach to Adapting Behavior Foundation Models
1
1
6
330
Ron Vainshtein retweeted
Parallels between AI and #LK99: - Both are modern-day alchemy. Just try more recipes until eureka. - The holy grail is simpler than we expected. - Lots of hyperparameters to tune. - Random seeds matter. - Arxiv is the new battlefield. - Scaling up is the key. - Tend to break the internet with every update. - Fastest growth in Twitter experts. - Democratization: there will be GPT-4's from big corps, and Alpaca's from grassroots.
32
137
15
945
309,502
Ron Vainshtein retweeted
ืœื™ื™ื‘ื˜ื•ื™ื˜ ืžื”ืฉื™ื—ื” ืฉืœ ืกืื ืืœื˜ืžืŸ ื‘ืื•ื ' ืชืœ ืื‘ื™ื‘. ืœื ื‘ืœื™ื™ื‘, ื›ื™ ืœื ื”ื™ื™ืชื™ ืฉื. ืจื•ืื” ืืช ื”ื”ืงืœื˜ื” ื•ื–ื•ืจืง ื›ืืŸ ืืช ื›ืœ ื”ืžื—ืฉื‘ื•ืช ืฉืœื™ ืคืœื•ืก ืชืžืœื•ืœ. ืื–: ืกื ืืœื˜ืžืŸ, ืžื ื›"ืœ OpenAI ื•ื”ืฉื ืฉื—ืœืงื›ื ืคื—ื•ืช ืžื›ื™ืจื™ื - ืื™ืœื™ื” ืกื•ืฆืงื‘ืจ, ื”ืžื“ืขืŸ ื”ืจืืฉื™ ืฉืœ OpenAI ืฉื”ื•ื ื™ืฉืจืืœื™ ืœืฉืขื‘ืจ ืฉืœืžื“ ืคื” ื‘ืืจืฅ. >> youtube.com/watch?v=mC-0XqTAโ€ฆ
9
4
3
98
25,855
Nice, love the concept ๐Ÿ“ฆ
Some of my early experiments with @runwayml GEN-1 Itโ€™s still early days of AI video - this tech will only get better. A whole new generation of filmmakers is gonna be able to make whatever they want on zero budget See below for my process
1
2
352
Really excited to see this paper released! Great work by the authors! Hoping to see it integrated into diffusers soon ๐Ÿคฉ @huggingface
Attend-and-Excite: Attention-Based Semantic Guidance for Text-to-Image Diffusion Models abs: arxiv.org/abs/2301.13826 project page: attendandexcite.github.io/Atโ€ฆ github: github.com/AttendAndExcite/Aโ€ฆ
1
3
34
13,355
โ€œA Simple Configuration Library is All You Needโ€ Happy to share pyrallis ๐Ÿ‰, a small project we've been working on. Pyrallis is a simple library, derived from simple-parsing and inspired by Hydra, for automagically creating configs from dataclasses. eladrich.github.io/pyrallis/
3
7
2
35
Ron Vainshtein retweeted
Just got accepted to #ICLR2023 as "notable-top-25%" (spotlight) ๐Ÿคฉ #NeRN
What if we thought of a pretrained NN's weights as a coordinate-value system? Then we could learn a neural representation for that system (just like folks have been doing for 3D scenes)!! And yes, it actually works! Check out our new work #NeRN
3
20
867
Ron Vainshtein retweeted
ืชื•ืดื› ื”ื”ืชืœื”ื‘ื•ืช ื”ืขืฆื•ืžื” [ื•ื”ืžื•ืฆื“ืงืช] ืกื‘ื™ื‘ Dall-E 2 ืฆืฆื• ืœื• ืฉื ื™ ืดืžืชื—ืจื™ืืด. ืื—ื“ ืžืื—ื•ืจื™ ื‘ื˜ื ื•ื‘ืœื™ ืงื•ื“ ืคืชื•ื—, ื‘ืฉื Midjourney ืฉื–ื›ื” ืœืชื”ื•ื“ื” ื™ื—ืกื™ืช ื’ื“ื•ืœื” ื‘ืขืฆืžื• - ื•ื”ืฉื ื™ ื•ื”ืžืงื•ืคื— ื”ื•ื Stable Diffusion. ืœืžื” ื”ื•ื ืžื™ื•ื—ื“? >>
6
3
4
73
new fan-made NeurIPS 2022 trailer:
50
498
126
2,967
Ron Vainshtein retweeted
Super excited to share our new work! On a quest to find good regularization for Meta RL with @AvivTamar1 and Gilad Adler. Paper:ย arxiv.org/abs/2206.10716 Code:ย github.com/zoharri/MBRL2 [1/7]
New meta RL theory, relating the number of training tasks to the effective dimension of task space. Key idea - density estimation + dimensionality reduction for learning a model of possible tasks.
1
6
1
19
Ron Vainshtein retweeted
The community should focus more on positive ways to encourage reviewers instead of finding new ways to scare them away ๐Ÿ˜•
CVPR: We welcome new students! Students: Cool! I just submitted my very first paper! CVPR: You now also need to review papers. Students: Not sure how to do that. Here is my attempt. CVPR: Your reviews are too unprofessional. You are now banned for two years. Students: ...
4
4
Ron Vainshtein retweeted
Do brains build generative models all the way down to pixel level? I happened to get woken up this morning just as I was scrutinizing a visual detail in the dream, which gave me a strong sense that it does. Previously I've been less sure. Anyone else try to debug?
258
94
16
1,644
Ron Vainshtein retweeted
The elegance of ML is the elegance of biology, not the elegance of math or physics. Simple gradient descent creates mind-boggling structure and behavior, just as evolution creates the awe inspiring complexity of nature.
What are the most elegant/beautiful ideas in ML? Feels like mathematicians & physicists often talk about aesthetics, but we very rarely do. Why?
58
364
69
2,639
Ron Vainshtein retweeted
8
16
2
189
What would a peer review process look like if it was designed today? ๐Ÿค” We strongly believe that peer review can be so much better for both authors and reviewers and we're super excited to share @peerxiv, our vision for a fresh view on peer review๐Ÿ’ก peerxiv.web.app/about
10
41
9
160