Congrats' François Charton and Julia Kempe !

Very Impressive work! It appears that Multi-Token Prediction (MTP) has a substantial impact for both pre-training and inference.

A good opportunity to (re)-read this great paper: arxiv.org/abs/2404.19737

Congrats to all this team !

A pure physics paper based on intuitions from AI experiments, expect more of these!

This is only the beginning ;)

An opportunity to work with amazing people :)

This "small group at FAIR Paris" was amazing 😉

Grande classe Mistral AI !

We ran 5,600 hyperparameter sweeps to compare RL algorithms on hidden-information games with billions of states. In our benchmark, we found that properly tuned policy gradient methods, such as PPO, performed the best. Paper: arxiv.org/abs/2502.08938

Last week I went to the AI & Maths conference organized by Université PSL 👨‍💻🤖 Quick thoughts:

A pleasure and an honor!

We believe formal math is the future. 🔥Introducing Kimina-Prover Preview, a Numina & Kimi.ai collaboration, the first large formal reasoning model for Lean 4, achieving 80.78% miniF2F. github.com/MoonshotAI/Kim…

Very impressive!

Hey OpenAI ! how can I get access API access to o3-mini ? I'm a Tier 5 user and it would be very helpful for my projects (in particular some research projects) !

Assistant professor level-position in machine learning in my university ! Funded by Hi! PARIS, and reduced teaching duties :) More info here: lnkd.in/er2Sfn_d. The application deadline is May 9. International applications are welcome!

Huge work with Zeyuan Allen-Zhu, Sc.D.: we build a synthetic arena to really understand the differences between model architectures and we introduce a simple but effective Canon layer to boost reasoning. Check it out!

So cool! I've been looking forward to it! It's also very cool and beneficial to the research community to have a document that explains everything in detail. Hats off to Mistral!

Un travail Magistral :)

Congratulations Jeremy! It's very good news to see this new institute, much needed in tomorrow's world.