Kirill Neklyudov (@k_neklyudov) / X

Kirill Neklyudov

335 posts

Kirill Neklyudov

@k_neklyudov

Member of the Technical Staff @AnthropicAI Assistant Professor @UMontreal (on leave) Core Academic Member @Mila_Quebec

Joined August 2016

Pinned
Kirill Neklyudov
@k_neklyudov
May 13
Population dynamics (eg murmuration of birds 🐦🐦🐦) is notoriously hard to learn; choosing the right model for the dynamics is even harder. In our #ICML2026 spotlight, we introduce Wasserstein Lagrangian Mechanics (WLM) for learning population dynamics from observations, which
GIF
24K
Kirill Neklyudov
@k_neklyudov
Jul 1, 2020
Our #icml2020 paper "Involutive MCMC: a Unifying Framework" is now available on arxiv arxiv.org/abs/2006.16653. It describes many MCMC algorithms from a single perspective. Work with @wellingmax, @eeevgen, Dmitry Vetrov
Kirill Neklyudov
@k_neklyudov
Nov 27, 2024
so happy to see that Action Matching finds its applications in physics, outperforming diffusion models and Flow Matching! wonderful work by Jules Berman, Tobias Blickhan, and Benjamin Peherstorfer! arxiv.org/abs/2410.12000
GIF
GIF
21K
Kirill Neklyudov
@k_neklyudov
Aug 29, 2024
Meta Flow Matching The evolution of many stochastic processes depends on the current distribution of samples. Indeed, for the diffusion, we can see that the vector field propagating the particles depends on the current density (through the score). It is also evident in the
Lazar Atanackovic
@lazar_atan
Aug 29, 2024
🚀Introducing — Meta Flow Matching (MFM) 🚀 Imagine predicting patient-specific treatment responses for unseen cases or building generative models that adapt across different measures. MFM makes this a reality. 📰Paper: arxiv.org/abs/2408.14608 💻Code: github.com/lazaratan/meta…
28K
Kirill Neklyudov
@k_neklyudov
Jun 24, 2025
(1/n) Sampling from the Boltzmann density better than Molecular Dynamics (MD)? It is possible with PITA 🫓 Progressive Inference Time Annealing! A spotlight @genbio_workshop of @icmlconf 2025! PITA learns from "hot," easy-to-explore molecular states 🔥 and then cleverly "cools"
50K
Kirill Neklyudov
@k_neklyudov
Dec 28, 2024
🧵(1/5) Have you ever wanted to combine different pre-trained diffusion models but don't have time or data to retrain a new, bigger model? 🚀 Introducing SuperDiff 🦹‍♀️ – a principled method for efficiently combining multiple pre-trained diffusion models solely during inference!
GIF
111K
Kirill Neklyudov
@k_neklyudov
Oct 14, 2024
Simulating protein folding between two metastable states is possible without any training data! Our #NeurIPS2024 spotlight paper brings us one step closer to this goal! arxiv.org/abs/2410.07974 This project would not see the light without @YuanqiD @MichaelPlainer @brekelmaniac
32K
Kirill Neklyudov
@k_neklyudov
Sep 8, 2024
The continuity equation: the very fibre of generative modelling video: youtu.be/uK-apwLuEk8?si… I'm explaining the intuitive way to derive this equation. This is important not only for understanding generative modelling but also for using it in different contexts. Also, you can
19K
Kirill Neklyudov
@k_neklyudov
Jul 6, 2024
Wasserstein Lagrangian Flows explain many different dynamics on the space of distributions from a single perspective. arxiv.org/abs/2310.10649 I made a video explaining our (with @brekelmaniac) #icml2024 paper about WLF. Like subscribe share, lol. youtu.be/kkddiLegc3s?si…
19K
Kirill Neklyudov
@k_neklyudov
Oct 25, 2023
Feel lost in all the kinds of optimal transport (OT)? Lagrangian flow is a single concept behind all of them. We (with @brekelmaniac) use it to propose a unified computational approach to - OT - Schrödinger bridge - OT with physical prior - unbalanced OT arxiv.org/abs/2310.10649
46K
Kirill Neklyudov
@k_neklyudov
Jul 21, 2023
Wasserstein Quantum Monte Carlo is a new method for the ground state estimation of a quantum system arxiv.org/abs/2307.07050 ML community might find this methodology interesting since it applies far beyond QM + @JannesNys @lucaathiede @carrasqu @lqiang67 @wellingmax @AliMakhzani
71K
Kirill Neklyudov
@k_neklyudov
Apr 17, 2024
Je vais à Montréal! This June I'm starting a new position as an assistant professor at @UMontreal and as a core academic member of @Mila_Quebec. Drop me a line if you're interested in working together on problems in AI4Science, Optimal Transport, and Generative Modeling.
12K
Kirill Neklyudov
@k_neklyudov
Jun 17, 2025
Why do we keep sampling from the same distribution the model was trained on? We rethink this old paradigm by introducing Feynman-Kac Correctors (FKCs) – a flexible framework for controlling the distribution of samples at inference time in diffusion models! Without re-training
Marta Skreta
@martoskreto
Jun 17, 2025
🧵(1/6) Delighted to share our @icmlconf 2025 spotlight paper: the Feynman-Kac Correctors (FKCs) in Diffusion Picture this: it’s inference time and we want to generate new samples from our diffusion model. But we don’t want to just copy the training data – we may want to sample
GIF
arxiv.org
Feynman-Kac Correctors in Diffusion: Annealing, Guidance, and...
While score-based generative models are the model of choice across diverse domains, there are limited tools available for controlling inference-time behavior in a principled manner, e.g. for...
30K
Kirill Neklyudov
@k_neklyudov
Jul 19, 2024
I'm going to be at #ICML2024. DM me if you want to chat, especially, if you would like to do a PhD or a Master's with me at @Mila_Quebec Me and my coauthors are presenting 5 works next week Main track: - A Computational Framework for Solving Wasserstein Lagrangian Flows
18K