I'm hiring at least one post-doc! We're interested in creating language models that process language more like humans than mainstream LLMs do, through architectural modifications and interpretability-style steering.

Wow, data pipeline is very important! Great work, congrats:P

Our latest CS336 Language Modeling from Scratch lectures are now available! View the entire playlist here: youtube.com/playlist?list=…

The Code of Practice is out. I co-wrote the Safety & Security Chapter, which is an implementation tool to help frontier AI companies comply with the EU AI Act in a lean but effective way. I am proud of the result! 1/3

This feeling catches me up when I saw Grok4's publishing video. I think currently its very hard to claim A model better than B, everyone should has her own private benchmark!

Join our Discord for discussions and a bunch of simple submission ideas you can try! discord.gg/n5uwjQcxPR Participants will have the option to write a system description paper that gets published.

🚨 Postdoc Hiring: I am looking for a postdoc to work on rigorously evaluating and advancing the capabilities and safety of computer-use agents (CUAs), co-advised with Yu Su OSU NLP Group. We welcome strong applicants with experience in CUAs, long-horizon reasoning/planning,

We're launching an "AI psychiatry" team as part of interpretability efforts at Anthropic!  We'll be researching phenomena like model personas, motivations, and situational awareness, and how they lead to spooky/unhinged behaviors. We're hiring - join us! job-boards.greenhouse.io/anthropic/jobs…

My Winter MATS applications are open! You'll work full-time writing a mech interp paper supervised by me. Due Aug 29 I've supervised 30+ papers by now (incl 15 top conference papers) but cohorts still get better each time. I'm hyped to see what this cohort achieves! Highlights:

I read the first 4 books extensively during my PhD, highly recommended 👍 I'd also highlight the 5th book as my first read re deep learning. Mind-blowing for a young math undergrad (me) at the time, made me decide to go for ML

Introducing Discord for research Discord was built for gamers. What about researchers? We’ve built a platform where researchers can discover communities, engage in paper discussions, and connect with academics who share your interests - all in one dedicated space.

🦥 Qwen3-Coder-Flash: Qwen3-Coder-30B-A3B-Instruct 💚 Just lightning-fast, accurate code generation. ✅ Native 256K context (supports up to 1M tokens with YaRN) ✅ Optimized for platforms like Qwen Code, Cline, Roo Code, Kilo Code, etc. ✅ Seamless function calling & agent

Congrats and thanks for sharing! Gonna try it out in my enterprise workflow.

I think we shouldn't regard Sutton's bitter lesson as the only Bible to achieve AGI but to have scaling as one dim in mind to explore while also research arch and "educate" recipe (data, curriculum, obj) and propose new paradigms, e.g. to subvert pre/mid/post train pipeline.

Introducing RynnEC, our little step towards physical world understanding🚀🚀🚀 1. RynnEC is object-centric, supporting the recognition of up to 12 object properties/relations. 2. RynnEC is space-aware using RGB videos only (45.8 on vsi-bench), no explicit 3D encoding required

Our new lab for Human & Machine Intelligence is officially open at Princeton University! Consider applying for a PhD or Postdoc position, either through the depts. of Computer Science or Psychology. You can register interest on our new website lake-lab.github.io (1/2)

Ever wonder what's really happening when we use RL to teach LLMs to reason? 🤔 The process is full of mysteries. 🤯 What causes those sudden "aha moments" in training? 📏 Why does better reasoning often lead to longer answers ("length-scaling")? 📉 Why does token entropy often

I’m recruiting PhD students for 2026! If you are interested in robustness, training dynamics, interpretability for scientific understanding, or the science of LLM analysis you should apply. BU is building a huge LLM analysis/interp group and you’ll be joining at the ground floor.

I recently changed my position with AISI (the UK govt's AI security institute) to an expert advisor position which allows me again to express my views publicly and engage with the media, and I have some stuff to share. I've been reading Sir Tim Berners-Lee's recently released

⛵Marin 32B Base (mantis) is done training! It is the best open-source base model (beating OLMo 2 32B Base) and it’s even close to the best comparably-sized open-weight base models, Gemma 3 27B PT and Qwen 2.5 32B Base. Ranking across 19 benchmarks: