Quentin Bertrand
@Qu3ntinB
Researcher at @Inria. Previously, postdoctoral researcher at @Mila_Quebec w/ @SimonLacosteJ and @gauthier_gidel.
AI4Mat is back for NeurIPS! time to crystallize those ideas and make a solid-state submission by august 22, 2025 💪 new this year: opt-in your work for our Research Learning from Speaker Feedback program -- a new structured discussion format where spotlight presenters receive…
All the reviews of all my NeurIPS submissions are of high quality. They're not all positive, but they're all courteous and constructive, and nothing looks LLM-generated.
Had an amazing time at #ICML2025. So many inspiring talks and thoughtful conversations throughout the week! I also gave a talk and got great questions and feedback. ▶️ For anyone interested, here is the recording: icml.cc/virtual/2025/s…
🧵 Everyone is chasing new diffusion models—but what about the representations they model from? We introduce Discrete Latent Codes (DLCs): - Discrete representation for diffusion models - Uncond. gen. SOTA FID (1.59 on ImageNet) - Compositional generation - Integrates with LLM 🧱
Wrote a new post about Flow Matching, where I try to explain my intuitions about it as concisely as possible. Hope it's useful :)
the first paper I linked opposes this: they show that in high-dim settings, if you run the denoising process even for a little bit, the average trajectories "clarifies" itself to a clear target from the training data. this is great insight!
the analysis here was a pleasure to read! they design experiments to tease out different possible factors that could lead to generalization in flow-matching, & argue that it's ultimately that capacity is what prevents the model from reproducing the boring empirical distributn
Yes! We precisely analyze this phenomenon in our recent paper: ¨On the Closed-Form of Flow Matching: Generalization Does Not Arise from Target Stochasticity¨ arxiv.org/pdf/2506.03719
Round of applause for our fantastic organizers: Sophia Abraham, Ninon Lizé Masclef, Mariya Hendriksen, and Meriem Mehri 🫶 #WiML #ICML2025
Our London team at @Xaira_Thera is looking for an ML/AI Research Engineer to build and scale antibody design 🦈🦙. Reach out if you enjoy working in an interdisciplinary team and want to push the boundaries of science (maybe 👀) job-boards.greenhouse.io/xairatherapeut…
If you're at #ICML2025 we are presenting our two papers on f-divergence generated losses and learning the log-partition of EBMs in the 11am poster session today. Posters W-508 and W-513 !
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.
NeurIPS is pleased to officially endorse EurIPS, an independently-organized meeting taking place in Copenhagen this year, which will offer researchers an opportunity to additionally present their accepted NeurIPS work in Europe, concurrently with NeurIPS. Read more in our blog…
we’re not kfc but come watch us cook with our feynman-kac correctors, 4:30 pm today (july 16) at @icmlconf poster session — east exhibition hall #3109 @k_neklyudov @AlexanderTong7 @tara_aksa @OhanesianViktor
I am at ICML, presenting our new convergence results for Scaffold with **linear speed-up**. It's poster W-621 between 1:30 and 4:00. Come and chat! Paper: arxiv.org/abs/2503.07594
Come check out SBG happening now! W-115 11-1:30 with @charliebtan @bose_joey Chen Lin @leonklein26 @mmbronstein
Happy to have presented the first open-vocabulary sound event detection training method that works, at #ICML2025, with @wuyusongwys! arxiv.org/abs/2505.05335
Hello #ICML2025👋, anyone up for a diffusion circle? We'll just sit down somewhere and talk shop. 🕒Join us at 3PM on Thursday July 17. We'll meet here (see photo, near the west building's west entrance), and venture out from there to find a good spot to sit. Tell your friends!
Autoencoder neural networks to learn Koopman eigenfunctions? Here's a great one: youtube.com/watch?v=74lLr_…
I wrote a post on how to connect with people (i.e., make friends) at CS conferences. These events can be intimidating so here's some suggestions on how to navigate them I'm late for #ICLR2025 #NAACL2025, but just in time for #AISTATS2025 and timely for #ICML2025 acceptances! 1/4