Jerry Zhi-Yang He

Is your robot 🤖 safe around humans🚸? Or will it fail catastrophically🤕? Excited to introduce "Natural Adversarial Frontier", a framework for probing the robustness of Human-Robot Interactions. To appear at #CoRL2023 🧵(1/8) Joint work with @ancadianadragan @daniel_s_brown @ZackoryErickson arxiv.org/abs/2310.10610 ood-human.github.io

Donald Knuth is vibemathing now. real tough day for the stochastic-parrot crew.

As promised, here's the short film Jia Zhangke produced using Seedance 2.0 for Chinese New Year and his take on AI filmmaking

As Rocks May Think: an interactive essay on thinking models, automated research, and where I think they are headed. Enjoy! evjang.com/2026/02/04/roc…

@DorsaSadigh Congrats Dorsa! Really happy for you!

Just realized I haven't shared life status or been on twitter for a while, so here is a status dump 🧵 1/4

The @ilyasut episode 0:00:00 – Explaining model jaggedness 0:09:39 - Emotions and value functions 0:18:49 – What are we scaling? 0:25:13 – Why humans generalize better than models 0:35:45 – Straight-shotting superintelligence 0:46:47 – SSI’s model will learn from deployment 0:55:07 – Alignment 1:18:13 – “We are squarely an age of research company” 1:29:23 – Self-play and multi-agent 1:32:42 – Research taste Look up Dwarkesh Podcast on YouTube, Apple Podcasts, or Spotify. Enjoy!

@WinstonGu_ very cool!

It would be really exciting if someone could build on this work for humanoid robots! Original code at github.com/Winston-Gu/Coo…

I have always been surprised by how few positive samples adversarial imitation learning needs to be effective. With ADD we take this to the extreme! A differential discriminator trained with a SINGLE positive sample can still be effective for a wide range of tasks.

I’ve also been integrating LLMs into my research workflow. I spent most of Tuesday working on a problem I’ve been thinking about for a while with some collaborators. I had a conjecture on a possible way forward, and with some hours of thinking, mixing in conversations with Gemini to guide certain non-trivial calculations, Gemini ultimately spit out a proof that no approach in this family can possibly work (which I found surprising, since similar approaches worked in related settings). Maybe will say more about what the problem is after it’s fully resolved, lest I lead to us getting scooped. :) tl;dr LLMs haven’t replaced me (yet?), but certainly are making me a more efficient researcher. *work still ongoing*

As a researcher at a frontier lab I’m often surprised by how unaware of current AI progress public discussions are. I wrote a post to summarize studies of recent progress, and what we should expect in the next 1-2 years: julian.ac/blog/2025/09/2…

What does it take to build a human-like user simulator? // To train collaborative agents, we need better user sims. In blog post pt 2, @NickATomlin and I sketch a framework for building user simulators + open questions for research: jessylin.com/2025/09/25/use…

@michellearning @medra_ai Welcome back!

The exciting news I never announced in 2021: I’m going to join NYU Courant as assistant professor! But that never happened because I decided to start @medra_ai instead 😅 Also hello world, I’m back on X four years later 👋

@YuXiang_IRVL @lfcasas7 @UT_Dallas This is super cool!

Big day in class today! With @lfcasas7, we brought 14 SO-101 arms for students to assemble and take home for projects @UT_Dallas. Excited to see what they create by semester’s end! 🤖

Today Thinking Machines Lab is launching our research blog, Connectionism. Our first blog post is “Defeating Nondeterminism in LLM Inference” We believe that science is better when shared. Connectionism will cover topics as varied as our research is: from kernel numerics to prompt engineering. Here we share what we are working on and connect with the research community frequently and openly. The name Connectionism is a throwback to an earlier era of AI; it was the name of the subfield in the 1980s that studied neural networks and their similarity to biological brains. thinkingmachines.ai/blog/defeating…

@robot_trainer Any particular failure mode of classical motion generation methods that you have in mind? Those seem to be pretty good & robust at reactive obstacle avoidance, if engineered well

this is really cool. i've always thought learning-based methods were the right approach to global motion generation. nice work! (and all the demos! super robust and general system)

Our IMO gold medal-winning AI pipeline is now model-agnostic. 🥇 What worked for Gemini 2.5 Pro now gets the same 5/6 score with GPT-5 & Grok4. This confirms the power of our verification-and-refinement pipeline to improve base model capabilities. The new code & results are live on GitHub[github.com/lyang36/IMO25]! Paper update coming soon. Huge thanks to @xai for the Grok4 API credits! #AI #LLM #IMO #MathOlympiad #OpenSource”

We are thrilled to introduce the Seed-OSS family of open-source LLMs, developed by ByteDance's Seed Team. GitHub: github.com/ByteDance-Seed… HuggingFace: huggingface.co/collections/By… Feel free to try it out and share your feedback!

@alexwei_ Will the solutions be posted on github similar to IMO results?

1/ I competed for Team USA at IOI in 2015, so this achievement hits home for me. The biggest highlight: we *did not* train a model specifically for IOI. Our IMO gold model actually set a new state of the art in our internal competitive programming evals. Reasoning generalizes!

🤖Introducing: CoT-Self-Instruct 🤖 📝: arxiv.org/abs/2507.23751 - Builds high-quality synthetic data via reasoning CoT + quality filtering - Gains on reasoning tasks: MATH500, AMC23, AIME24 & GPQA-💎 - Outperforms existing train data s1k & OpenMathReasoning - Gains on non-reasoning tasks as well: AlpacaEval & ArenaHard 🧵1/3

looks like Lean is popular today, so here's a little post on how/why it works and implementing a mini version of it in Julia

@conitzer when you click on "dive deeper in AI mode" it seems to be doing fine

"is it possible to swim in coffee?" -- apparently it is not because coffee is a liquid, not a solid, and you want a solid, swimmable medium.