We need innovative technical and societal solutions to mitigate AI risks. I believe liability insurance for AI developers could be an excellent market-based incentive to drive safety standards and accountability, and is an option worth considering. https://t.co/SXD1pRSSz1

We just shipped a major Mercury refresh. ⚡ Best-in-class quality at up to 10× lower latency. Still the only commercial diffusion LLM in the world. Try the new model.

This Stanford professor just raised a $50M Seed and have built a 10x faster and 10x cheaper AI coding model with the performance of Gemini Flash / Haiku. Inception Labs’ Mercury model can implement games like Connect 4 from scratch in ~2s. The speed feels magical, like going

Being part of Mila means joining the world’s largest community of academic researchers in deep learning. Submit your supervision request now for the MSc or PhD for Fall 2026. https://t.co/r01eLcXtZw

Very grateful to @schmidtsciences for being awarded an #AI2050 senior fellowship. And honored to be part of this select 2025 cohort of 7 senior fellows. This award will support our work on a deeper scientific basis for understanding and improving how artificial intelligence

Why and how does gradient/matrix orthogonalization work in Muon for training #LLMs? We introduce an isotropic curvature model to explain it. Take-aways: 1. Orthogonalization is a good idea, "on the right track". 2. But it might not be optimal. [1/n]

Come do a PhD with me 😀! Promise of fun science and great coffee ☕

🌀New paper on the generation phases of Flow Matching https://t.co/tzG2kPVGsE Are FM & diffusion models nothing else than denoisers trained at every noise level? In theory yes, *if trained optimally*. But in practice, do all noise level matter equally?

🚨🚨 The common myth that value alignment happens at the preference optimization (RLHF) stage is incorrect and have mislead years of research 💣. Mehar did a meticulous job showing that LLMs acquire values during SFT, not during preference optimization. Your SFT is probably the

Pushed a big update to LM-class (v2025.2) -- this second version makes a much more mature resource Many refinements of lecture slides + significant improvements to the assignments Many thanks to @ch272h @HuaYilun and @shankarpad8 for their work on the assignments

Full stack devs, SWEs, MLEs, forward deployed engineers, research engineers, applied scientists: we are hiring! Join us and tackle cutting-edge challenges including physical AI, time series, material sciences, cybersecurity and many more. Positions available in Paris, London,

I contemplated whether I should post this, because it seems kind of obvious. But it's often taken for granted, so we might underestimate the impact: e.g. these days, diffusion papers don't usually show samples without guidance anymore (figures from GLIDE https://t.co/2wOdFfRHCK)

🚨In our NeurIPS paper, we bring encoder-decoders back.. for diffusion language models! ⚡️Encoder-decoders make diffusion sampling fast: a small (fast) decoder denoises tokens progressively and a large (slower) encoder represents clean context.

Recordings for the talk on "Tiny Recursive Models" are up. https://t.co/6iCRHcX4VL

Introducing Generalised Flow Maps 🎉 A stable, few-step generative model on Riemannian manifolds 🪩 📚 Read it at: https://t.co/iCTHedwCxf 💾 Code: https://t.co/MeukcthFN2 @msalbergo @nmboffi @mmbronstein @bose_joey

Preprint on using ChatGPT to resolve a 42-year-old open problem (point convergence of Nesterov’s accelerated gradient method) is out. Mathematical results are complete, though still need to expand the discussion of historical context & prior work. (1/2) https://t.co/Dmd9huMjXS

I am potentially recruiting a postdoctoral fellow through this program. If interested, name me as a mentor, and ping me to let me know that you are applying! The process includes some sort of interview, so I can try to squeeze a few of these in advance (it will help a lot!)

Deep Image Prior is back

I will post the recordings after, don't worry.

I will give a presentation on "Tiny Recursion Models" tomorrow at 1pm in the Mila Agora (6650 Saint-Urbain, Montreal). Its open to everyone, feel free to come by!