Mathieu Blondel
mblondel.bsky.social
Mathieu Blondel
@mblondel.bsky.social
Research scientist, Google DeepMind
Reposted by Mathieu Blondel
📣 Please share: We invite submissions to the 29th International Conference on Artificial Intelligence and Statistics (#AISTATS 2026) and welcome paper submissions at the intersection of AI, machine learning, statistics, and related areas. [1/3]
August 12, 2025 at 11:46 AM
Reposted by Mathieu Blondel
I'm not on TV yet, but I'm on YouTube 😊 talking about research, ML, how I prepare talks and the difference between Bayesian and frequentist statistics.

Many thanks to Charles Riou who already posted many videos of interviews of ML & stats researchers on his YouTube channel "ML New Papers"!! 🙏
April 28, 2025 at 11:08 AM
Reposted by Mathieu Blondel
1.5 yrs ago, we set out to answer a seemingly simple question: what are we *actually* getting out of RL in fine-tuning? I'm thrilled to share a pearl we found on the deepest dive of my PhD: the value of RL in RLHF seems to come from *generation-verification gaps*. Get ready to 🤿:
March 4, 2025 at 8:59 PM
Schrödinger's snack
There is a trace of someone's sadness.
February 18, 2025 at 8:43 AM
Reposted by Mathieu Blondel
🧗‍♂️Why GD converges beyond [step size]<2/[smoothness]? We investigate loss functions and identify their *separation margin* is an important factor. Surprisingly Renyi 2-entropy yields super fast rate T=Ω(ε^{-1/3})!
arxiv.org/abs/2502.04889
February 10, 2025 at 3:04 AM
Reposted by Mathieu Blondel
Modern post-training is essentially distillation then RL. While reward hacking is well-known and feared, could there be such a thing as teacher hacking? Our latest paper confirms it. Fortunately, we also show how to mitigate it! The secret: diversity and onlineness! arxiv.org/abs/2502.02671
February 8, 2025 at 6:06 PM
The EBM paper below parameterizes dual variables as neural nets. This idea (which has been used in other contexts such as OT or GANs) is very powerful and may be *the* way duality can be useful for neural nets (or rather, neural nets can be useful for duality!).
Really proud of these two companion papers by our team at GDM:

1) Joint Learning of Energy-based Models and their Partition Function
arxiv.org/abs/2501.18528

2) Loss Functions and Operators Generated by f-Divergences
arxiv.org/abs/2501.18537

A thread.
January 31, 2025 at 1:53 PM
Really proud of these two companion papers by our team at GDM:

1) Joint Learning of Energy-based Models and their Partition Function
arxiv.org/abs/2501.18528

2) Loss Functions and Operators Generated by f-Divergences
arxiv.org/abs/2501.18537

A thread.
January 31, 2025 at 12:06 PM
Reposted by Mathieu Blondel
Sparser, better, faster, stronger
You think Jacobian and Hessian matrices are prohibitively expensive to compute on your problem? Our latest preprint with @gdalle.bsky.social might change your mind!
arxiv.org/abs/2501.17737
🧵1/8
January 30, 2025 at 3:57 PM
Former French minister of Education and "philosopher" Luc Ferry, who said a few years ago that maths was useless, wrote a book on artificial intelligence 😂
January 27, 2025 at 11:57 AM
Reposted by Mathieu Blondel
We are organising the First International Conference on Probabilistic Numerics (ProbNum 2025) at EURECOM in southern France in Sep 2025. Topics: AI, ML, Stat, Sim, and Numerics. Reposts very much appreciated!

probnum25.github.io
November 17, 2024 at 7:06 AM
Reposted by Mathieu Blondel
Slides for a general introduction to the use of Optimal Transport methods in learning, with an emphasis on diffusion models, flow matching, training 2 layers neural networks and deep transformers. speakerdeck.com/gpeyre/optim...
January 15, 2025 at 7:08 PM
Reposted by Mathieu Blondel
MLSS coming to Senegal !

📍 AIMS Mbour, Senegal
📅 June 23 - July 4, 2025

An international summer school to explore, collaborate, and deepen your understanding of machine learning in a unique and welcoming environment.
Details: mlss-senegal.github.io
MLSS Senegal 2025
mlss-senegal.github.io
December 12, 2024 at 8:45 AM
Reposted by Mathieu Blondel
Thrilled to be co-organizing NeurIPS in Paris at @sorbonne-universite.fr next week!

📑 100 papers from NeurIPS 2024. Nearly twice as many as in 2023!
🧑‍🎓 over 300 registered participants
✅ a local and sustainable alternative to flying to Vancouver.

More info: neuripsinparis.github.io/neurips2024p...
4th and 5th of December Sorbonne Center for Artificial Intelligence (SCAI)
neuripsinparis.github.io
November 27, 2024 at 4:09 PM
Why would blue sky be more successful than mastodon at replacing twitter for the scientific community? (honest question)
November 23, 2024 at 5:52 PM