Matthieu Lin @capybara_ai
PhD@TsinghuaCS Interested in Reinforcement Learning, LLM-based Agents, Alignment. linyuhongg.github.io Joined December 2023-
Tweets178
-
Followers75
-
Following994
-
Likes425
Interesting findings! We also attempted something similar in our AZR paper section D.2, where the proposer needs to construct a composite function f(g,..g)
Interesting findings! We also attempted something similar in our AZR paper section D.2, where the proposer needs to construct a composite function f(g,..g)
Introducing 🛡️ExCyTIn‑Bench: Evaluating LLM agents on Cyber Threat Investigations. It’s built on Azure tenant, a real Security Operations Center environment, covering 57 tables. Explore how LLMs fare in realistic, multi-hop incident detection! #Cybersecurity #AI #LLM #Benchmark
People love 𝗽𝗮𝘀𝘀@𝗸 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴. What to do if you have 𝟭𝟬𝟬 samples and you wanna optimize 𝗽𝗮𝘀𝘀@𝟭𝟬? ✨This is the reward. Prsented in the analytic form. Next step? Pass it to GRPO and witness the magic.
People love 𝗽𝗮𝘀𝘀@𝗸 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴. What to do if you have 𝟭𝟬𝟬 samples and you wanna optimize 𝗽𝗮𝘀𝘀@𝟭𝟬? ✨This is the reward. Prsented in the analytic form. Next step? Pass it to GRPO and witness the magic. https://t.co/D8KM0rQm1s
Today we're putting out an update to the JAX TPU book, this time on GPUs. How do GPUs work, especially compared to TPUs? How are they networked? And how does this affect LLM training? 1/n
getting a phd is mostly about developing a system for maximizing the time you spend in flow state, over several years
🔍We introduce ASearcher, a search agent trained by end2end RL Large-scale (up to 128 turns) RL with AReaL unlocks Long-Horizon Agentic Search (+20.8/+46.7% on GAIA/xBench) 💻Data, Code&Model: github.com/inclusionAI/AS… 📄Paper: arxiv.org/abs/2508.07976v #Agent #OpenSource #LLM #AGI
OpenAI hasn’t open-sourced a base model since GPT-2 in 2019. they recently released GPT-OSS, which is reasoning-only... or is it? turns out that underneath the surface, there is still a strong base model. so we extracted it. introducing gpt-oss-20b-base 🧵
[0/3] 🚀 Introducing Verlog – an open-source RL framework built specifically for training long-horizon, multi-turn LLM agents. 📊 Max episode length comparison: •VeRL / RAGEN → ~10 turns •verl-agent → ~50 turns •Verlog (ours) → 400+ turns 🔥 ⚙️ Technical foundation:…
⚡𝐅𝐏𝟖 makes RL faster — but at the cost of performance. We present 𝐅𝐥𝐚𝐬𝐡𝐑𝐋, the first 𝐨𝐩𝐞𝐧–𝐬𝐨𝐮𝐫𝐜𝐞 & 𝐰𝐨𝐫𝐤𝐢𝐧𝐠 𝐑𝐋 𝐫𝐞𝐜𝐢𝐩𝐞 that applies 𝐈𝐍𝐓𝟖/𝐅𝐏𝟖 for rollout 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐥𝐨𝐬𝐢𝐧𝐠 𝐩𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞 compared to 𝐁𝐅𝟏𝟔! 📝 Blog:…
I think I found @merettm's favorite t-shirt
Failing on 𝐥𝐚𝐫𝐠𝐞-𝐬𝐜𝐚𝐥𝐞 𝐑𝐋 with VeRL? ⚠️ Mixing inference backend (𝐯𝐋𝐋𝐌/𝐒𝐆𝐋𝐚𝐧𝐠) with training backends (𝐅𝐒𝐃𝐏/𝐌𝐞𝐠𝐚𝐭𝐫𝐨𝐧) 𝐬𝐞𝐜𝐫𝐞𝐭𝐥𝐲 𝐭𝐮𝐫𝐧𝐬 𝐲𝐨𝐮𝐫 𝐑𝐋 𝐢𝐧𝐭𝐨 𝐨𝐟𝐟-𝐩𝐨𝐥𝐢𝐜𝐲 — even if they share the same weights! 📉 Blog:…
Today we’re releasing a prototype of Genesys, an autonomous multi-agent LLM discovery system that aims to discover new types of language model architectures. We found Genesys can discover novel architectures competitive with the industry-standard transformer. 🧵
OMW to #ICML2025, reach out (on X or Whova) if you’re interested in talking about RL, reasoning/safety of LLMs, agents. I will also be presenting our AI4MATH workshop paper: limits-of-RLVR, 13:45-14:00 pm, July 18, 2025. Location: Ballroom C, feel free to drop by
Kimi K2 is here! The first big beautiful model purpose-built for agentic capabilities is now open-source! Agent RL, ready for takeoff!
Kimi K2 is here! The first big beautiful model purpose-built for agentic capabilities is now open-source! Agent RL, ready for takeoff!
Our first End2End RL Trained Agent is out! Hope you like it!
Our first End2End RL Trained Agent is out! Hope you like it!
Q-learning is not yet scalable seohong.me/blog/q-learnin… I wrote a blog post about my thoughts on scalable RL algorithms. To be clear, I'm still highly optimistic about off-policy RL and Q-learning! I just think we haven't found the right solution yet (the post discusses why).
I just had a dream where I wasnt fat
Thoughts on Two Papers: 1. our paper Limit-of-RLVR: Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model? 2. ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models See following points:
🚨Beyond 80/20 in LLM reasoning🚨Dropping 80% low-entropy tokens in RL greatly boosts performance 🔗arxiv.org/abs/2506.01939 🏆Zero-RL SoTA: 63.5/68.1 (AIME24), 56.7 (AIME25) 🚀Insights: 1. RL retains base model entropy patterns 2. High-entropy tokens drive all RL improvement ⬇️
🚨 Your RL only improves 𝗽𝗮𝘀𝘀@𝟭, not 𝗽𝗮𝘀𝘀@𝗸? 🚨 That’s not a bug — it’s a 𝗳𝗲𝗮𝘁𝘂𝗿𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗼𝗯𝗷𝗲𝗰𝘁𝗶𝘃𝗲 you’re optimizing. You get what you optimize for. If you want better pass@k, you need to optimize for pass@k at training time. 🧵 How?
Asuka🎀Redpanda @VoidAsuka
20K Followers 4K Following senior sde, junior ai research engineer. share something good/fun.
Robert Scoble @Scobleizer
539K Followers 17K Following The best from AI community | Ex-Microsoft, Rackspace, Fast Company | Wrote eight books about the future | Silicon Valley robots, holodecks, BCIs, and startups.
Auwuieadtou @Auwuieadtou548
29 Followers 1K Following
Asuka小能猫 @AsukaOdysseus
9K Followers 575 Following senior SDE, AI junior researcher. 颜值技术情感博主。 想阻止AI毁灭我的生活和我眷恋的世界。 小红书：薛定谔小原asuka 英推: @VoidAsuka ig：roninkamuy
Philippe Laban @PhilippeLaban
1K Followers 692 Following Research Scientist @MSFTResearch. NLP/HCI Research.
Johan Obando-Ceron �... @johanobandoc
2K Followers 4K Following Graduate student @Mila_Quebec @UMontrealDIRO | RL/Deep Learning/AI | De Cali/Colombia pal’ Mundo 🇨🇴 | #JuntosProsperamos⚡#TogetherWeThrive| 🌱🌎
Roberta @Roberta0923273
245 Followers 3K Following
Aqauqwar @Aqauqwar929992
1 Followers 140 Following
Feng Yao @fengyao1909
1K Followers 634 Following Ph.D. student @UCSD_CSE | Intern @Amazon Rufus Foundation Model Ex. @MSFTResearch @TsinghuaNLP
Ofir Press @OfirPress
15K Followers 6K Following I build tough benchmarks for LMs and then I get the LMs to solve them. SWE-bench & SWE-agent. Postdoc @Princeton. PhD @nlpnoah @UW.
Yiran Wu @YiranWu18
233 Followers 302 Following Co-Creator of #AutoGen #AG2, PhD Candidate @PennState. Intern @MSFTResearch 24 &25. Interested in Agentic AI, LLM reasoning and RL.
Kunhao Zheng @KunhaoZ
748 Followers 613 Following The real AGI is the friends we make along the way. PhD in FAIR CodeGen @AIatMeta. Alumni: @Huggingface, Sea AI Lab, @openai, École Polytechnique, SJTU
Cameron Cooper @ai_cats_chess
9 Followers 229 Following AI Explorer | Cat Lover 😼 | Chess Enthusiast ♟️
Aakanksha Chowdhery @achowdhery
11K Followers 5K Following @Stanford @reflection_ai // Previously @GoogleDeepMind :: PaLM, Gemini // @MSFTResearch, @Princeton // views my own and subject to change
Mikhail Doroshenko @SandelloRed
35 Followers 595 Following
Quantum Daddy @LegalPrimes
1K Followers 5K Following ⚛️🔧🥷🏽🦄 Financial primitives & exotic derivatives Founder @OG_Financial
Tianle Cai @tianle_cai
8K Followers 5K Following ML PhD @Princeton. Life-long learner, hacker, and thinker. Tech consultant & angel investor. Prev @togethercompute @GoogleDeepMind @MSFTResearch @citsecurities.
zirui @zirui3
45 Followers 2K Following Simplify hard & right things | Long-term focus | Efficient scaling🚀
Qihang Zhang @QihangZhang00
136 Followers 689 Following Ph.D student @UBC Prev: BEng @Tsinghua_Uni. Broadly interested in everything about LLMs. Try to build in public and learn in public.
Fforghysh @FforghyshKaBrI
29 Followers 2K Following A thousand-li journey is started by taking the first step.
Tacimus Sino @NonImperium
10 Followers 21 Following
: @YElo27154
6 Followers 1K Following
Rui Lu @RayLu_THU
332 Followers 136 Following PhD student in @Tsinghua_Uni studying machine learning theory, graduate from Yao class. Also a youtuber @ 漫士沉思录 manshi_math
a @yihyunCS
2K Followers 4K Following schizo-shitposter the child asked me if god exists. i said 'not yet.' prev @nasa
Theo @Pkjesuissurtwit
12 Followers 75 Following working to make my dad proud. I can't control everything, but my self-belief and my self-love are up to me. AI enthusiast.
Toughsh @ToughshEIr
4 Followers 118 Following
AISecHub @AISecHub
4K Followers 4K Following 🚀 AISecHub | AI & Cybersecurity | Discussing AI-driven threats, securing AI systems, and sharing insights on emerging challenges 💡
Gpbhupinder @gpbhupinder
469 Followers 7K Following 👨💻 Full-Stack Developer & AI Integration Expert 🚀 From concept to launch, we bring your tech vision to life
Tianzhu Ye @ytz2024
262 Followers 371 Following Foundation models. Student @Tsinghua_Uni. Research intern @MSFTResearch Asia.
Eldon Edwards @eldonredwards
3K Followers 8K Following Business consultant and popular culture and music researcher. Artificial intelligence researcher
Tianyu Pang @TianyuPang1
1K Followers 311 Following 🇸🇬Research Scientist at Sea AI Lab @SeaGroup; 👨🏻🎓PhD/BS from @Tsinghua_Uni and ex-@MSFTResearch; 🛡️Trustworthy AI and Generative Models.
Casper Hansen @casper_hansen_
10K Followers 457 Following NLP Scientist | AutoAWQ Creator | Open-Source Contributor
Qisen (Keysa) Yang @_KeysaYang
12 Followers 113 Following AI & ML researcher. Last year PhD student @Tsinghua_Uni.
Ying Sheng @ying11231
12K Followers 716 Following @lmsysorg | Prev. @xAI @Stanford | Assist Prof @UCLA. (Fall 2026) | Do it anyway | Live to fight another day
Mercor @mercor_ai
10K Followers 20 Following We use AI to understand human ability and match talent with the opportunities they're best suited for.
Zhihu Frontier @ZhihuFrontier
762 Followers 73 Following 🚀Bringing China's AI & tech trends, voices, and perspectives to the global stage. ⚡️Powered by Zhihu/知乎, China's leading knowledge community.
Ken Liu @kenziyuliu
2K Followers 875 Following CS PhD @StanfordAILab @StanfordNLP w/ @percyliang @sanmikoyejo. Past: DeepMind, CMU, USydney 🇦🇺
Sean Kirmani @SeanKirmani
3K Followers 547 Following Research @OpenAI. Interested in intelligence, understanding, and science.
马东锡 NLP @dongxi_nlp
14K Followers 762 Following Prev. PhD @Stockholm_Uni | Alumni @KTHuniversity @uppsalauni Sharing insights on AI, autonomous agents, and large language & reasoning models
Asuka🎀Redpanda @VoidAsuka
20K Followers 4K Following senior sde, junior ai research engineer. share something good/fun.
Pavel Durov @durov
2.7M Followers 1 Following Founder, CEO at @telegram (2013), founder, ex-CEO of @vkontakte (2006), part-time troll.
Chao Huang @huang_chao4969
2K Followers 99 Following Leading Data Intelligence Lab @HKUniversity | Recent Work: LightRAG, AutoAgent, RAG-Anything, AI-Researcher, DeepCode, MiniRAG, VideoRAG
Chris 🇨🇦 @llm_wizard
1K Followers 480 Following Working on cool open-source AI stuff @ NVIDIA Views my own.
gensyn @gensynai
85K Followers 36 Following Gensyn is building the network for machine intelligence. Build with us: https://t.co/U6bMeo46wM
Jackson Stokes @jackson_stokes
648 Followers 545 Following give me a break I’m on my journey . . . @trainloop_ai, ex XLA, Google Research
Cheng Lu @clu_cheng
8K Followers 200 Following Member of technical staff @OpenAI. PhD @Tsinghua_Uni. Interested in scalable generative models.
Laurens van der Maate... @lvdmaaten
4K Followers 2K Following Member of Technical Staff at Anthropic. Ex-Meta. t-SNE. Llama 3. DenseNet. Web-scale weakly supervised vision. CrypTen.
Jiayi Weng @Trinkle23897
3K Followers 142 Following MTS @openai, author of the entire post-training RL infra, core contributor of ChatGPT/GPT4/GPT4o etc. 30U30
LMSYS Org @lmsysorg
8K Followers 167 Following Large Model Systems Organization: Join our Slack: https://t.co/mSPNyKTLTS We developed SGLang https://t.co/jEqIJcGwGA, Chatbot Arena (now @lmarena_ai), and Vicuna!
Terry Yue Zhuo @ SF �... @terryyuezhuo
2K Followers 643 Following BigCodeBench: https://t.co/EYXhp0T8E5 | Now building beyond code | Hosting @codelm_tutorial at EMNLP’25
Gradient @Gradient_HQ
809K Followers 45 Following Building the world’s first fully distributed AI runtime on @solana.
Cerebras @CerebrasSystems
35K Followers 255 Following The world's fastest AI inference and training. Try the latest open models at: https://t.co/jREGhLI2nj
Yunhao (Robin) Tang @robinphysics
1K Followers 729 Following Interested in RL. Science @MistralAI. Prev Llama post-training @AIatMeta, Gemini post-training and deep RL research @Deepmind, PhD @Columbia
Yifei Wang @yifeiwang77
2K Followers 2K Following Postdoc @MIT_CSAIL. Self-supervised learning. Foundation Models. AI Safety. Prior BS+BA+PhD @PKU1898.
Banghua Zhu @BanghuaZ
7K Followers 1K Following Assistant Professor @UW, Principal Research Scientist @Nvidia. Prior Cofounder @NexusflowX, @Berkeley_EECS @Google @Microsoft. I work on LLMs.
Asuka小能猫 @AsukaOdysseus
9K Followers 575 Following senior SDE, AI junior researcher. 颜值技术情感博主。 想阻止AI毁灭我的生活和我眷恋的世界。 小红书：薛定谔小原asuka 英推: @VoidAsuka ig：roninkamuy
X. Dong @SimonXinDong
899 Followers 385 Following Research Scientist@NVIDIA . Making LLMs e.g., Hymba, Nemotron serials. Ex @Harvard @Meta @Tencent｜ Views and opinions are my own
Noah MacCallum @noahmacca
4K Followers 2K Following Applied AI @openai 🛠️ , ex-founder, engineer, bio researcher I care about safe deployment of AI, and applying it to advance science and improve human health.
Kiran Tomlinson @kiran_tomlinson
313 Followers 144 Following Researcher @ Microsoft Research. interested in choices, voting, preferences, LLMs, ….
Philippe Laban @PhilippeLaban
1K Followers 692 Following Research Scientist @MSFTResearch. NLP/HCI Research.
Calvin French-Owen @calvinfo
15K Followers 492 Following Making things, trail running. Prev: Codex @OpenAI, https://t.co/4qWGncHOAX, co-founder @Segment, @MIT
Paul Zhou @zhiyuan_zhou_
746 Followers 276 Following phd @berkeley_ai, research intern @physical_int
Databricks Mosaic Res... @DbrxMosaicAI
41K Followers 120 Following We remove the barriers to state-of-the-art generative AI model development and make data + AI available to all.
Andrea Michi @andreamichi
2K Followers 1K Following Co-founder @ https://t.co/FiVtWkCxXC / Building intelligence to detect and remediate software vulnerabilities / Prev post-training / RL for Gemini @GoogleDeepMind
Sara Hooker @sarahookr
49K Followers 9K Following I lead @Cohere_Labs. Formerly Research @Google Brain @GoogleDeepmind. ML Efficiency at scale, LLMs, ML reliability. Changing spaces where breakthroughs happen.
Dimitris Papailiopoul... @DimitrisPapail
20K Followers 1K Following Researcher @MSFTResearch, AI Frontiers Lab; Prof @UWMadison (on leave); learning in context; thinking about reasoning; babas of Inez Lily.
Haibin (on leave) @eric_haibin_lin
2K Followers 977 Following On the path to AGIs. Bytedance Seed LLM Systems, @verl_project, Megascale. Prev. @AmazonScience @ApacheMXNet
TRAE @Trae_ai
27K Followers 26 Following The Real AI Engineer. Download now: https://t.co/kapduuwttm Join SOLO waitlist: https://t.co/UXvS3gZnw7
rohan anil @_arohan_
25K Followers 2K FollowingTrends for United States
You might like
