Mathurin Massias

The Generation Phases of Flow Matching: a Denoising Perspective Flow matching has achieved remarkable success, yet the factors influencing the quality of its generation process remain poorly understood. In this work, we adopt a denoising perspective and design a f...

We dig into this equivalence in our latest preprint with @annegnx.bsky.social ! arxiv.org/abs/2510.24830

Strong afternoon session: Ségolène Martin on how to go from flow matching to denoisers (and hopefully come back?) and Claire Boyer on how learning rate and working in latent spaces affect diffusion models

Followed by Scott Pesme on how to use diffusion/flow matching based MMSE to compute a MAP (and nice examples!), and Thibaut Issenhuth on new ways to learn consistency models
@skate-the-apple.bsky.social

Next is @annegnx.bsky.social presenting our neurips paper on why flow matching generalizes, while it shouldn't!

arxiv.org/abs/2506.03719

Kickstarting our workshop on Flow matching and Diffusion with a talk by Eric Vanden Eijnden on how to optimize learning and sampling in Stochastic Interpolants!

Broadcast available at gdr-iasis.cnrs.fr/reunions/mod...

My paper on Generalized Gradient Norm Clipping & Non-Euclidean (L0, L1)-Smoothness (together with collaborators from EPFL) was accepted as an oral at NeurIPS! We extend the theory for our Scion algorithm to include gradient clipping. Read about it here arxiv.org/abs/2506.01913

merci David !

merci !

Our work on the generalization of Flow Matching got an oral at Neurips !

Go see @quentinbertrand.bsky.social present it there :)

🔥 Excited to announce the Workshop on the Principles of Generative Models at @euripsconf.bsky.social (the European conference parallel to NeurIPS 2025)
🇩🇰 Dec 6–7, Copenhagen
📝 Deadline for contributions: Oct 17
🔗 Website: sites.google.com/view/prigm-e...

Félicitations Anna !!

Oui, tout sera en anglais !

Oui !

Oui... c'est un compromis avec le fait d'avoir suffisamment de créneaux et de temps de discussions aux posters. Tu peux éventuellement arriver un peu après le début, et sinon ça devrait être accessible à distance 🤞

One-day workshop on Diffusion models and Flow matching, October 24th at @ensdelyon.bsky.social

Registration and call for contributions (short talk and poster) are open at
gdr-iasis.cnrs.fr/reunions/mod...

Super elegant approach !

on second thoughts I'm not sure I understood. In the classical FM loss you do have to learn this derivative no ? The loss is :

I was thinking of this:

I was thinking of the linear interpolant yes ; I haven't seen papers where other are used

it could even be velocity matching, and this time you do learn match the *conditional* velocities

Thanks for the kind words

Then why does flow matching generalize?? Because it fails!

The inductive bias of the neural network prevents from perfectly learning u* and overfitting.

In particular neural networks fail to learn the velocity field for two particular time values.

See the paper for a finer analysis 😀

We propose to regress directly against the optimal (deterministic) u* and show that it never degrades the performance
On the opposite, removing target stochasticity helps generalizing faster.

Yet flow matching generates new samples!

An hypothesis to explain this paradox is target stochasticity: FM targets the conditional velocity field ie only a stochastic approximation of the full velocity field u*

*We refute this hypothesis*: very early, the approximation almost equals u*

New paper on the generalization of Flow Matching www.arxiv.org/abs/2506.03719

🤯 Why does flow matching generalize? Did you know that the flow matching target you're trying to learn *can only generate training points*?

w @quentinbertrand.bsky.social @annegnx.bsky.social @remiemonet.bsky.social 👇👇👇

PnP-Flow: Plug-and-Play Image Restoration with Flow Matching In this paper, we introduce Plug-and-Play (PnP) Flow Matching, an algorithm for solving imaging inverse problems. PnP methods leverage the strength of pre-trained denoisers, often deep neural networks...

On Saturday Anne will also present some very, very cool work on how to leverage Flow Matching models to obtain sota Plug and Play methods:

PnP-Flow: Plug-and-Play Image Restoration with Flow Matching, poster #150 in poster session 6, Saturday at 3 pm

arxiv.org/abs/2410.02423

It was received quite enthusiastically here so time to share it again!!!

Our #ICLR2025 blog post on Flow Matching was published yesterday : iclr-blogposts.github.io/2025/blog/co...

My PhD student @annegnx.bsky.social will present it tomorrow in ICLR, 👉poster session 4, 3 pm, #549 in Hall 3/2B 👈

The proximal operator generalizes projection in convex optimization. It converts minimisers to fixed points. It is at the core of nonsmooth splitting methods and was first introduced by Jean-Jacques Moreau in 1965. www.numdam.org/article/BSMF...

🚢🚢 deepinv v0.3.0 is here, with many new features! 🚢 🚢

Our passionate team of contributors keeps shipping more exciting tools!

Deepinverse (deepinv.github.io) is a library for solving imaging inverse problems with deep learning.

I had a blast giving a summer school on generative models at AI Hub Senegal, in particular flow matching, with @quentinbertrand.bsky.social and @remiemonet.bsky.social

Our material is publicly available !!! github.com/QB3/SenHubIA...

ensdelyon.bsky.social