Steering Diffusion Models to Rare Events with Sequential Monte Carlo
Aavash Subedi, Tim Reichelt, Christopher Williams, Philip Stier, Yee Whye Teh, Saifuddin Syed
arxiv.org/abs/2610.08652 [𝚜𝚝𝚊𝚝.𝙼𝙻 𝚌𝚜.𝙻𝙶]
@NeurIPSConf 2026, here we come! 🚀
The @OxfordYss is thrilled to celebrate 12 main-track papers accepted from our members and their supervisors! 🎉
Many of us will be in Sydney 🇦🇺, Paris 🇫🇷 and Atlanta 🇺🇸 — come say hi!
I'm at @ISBA_events in Nagoya this week. Come say hi and check out our work!
🔶Metropolis Adjusted Diffusion Models
Wed (poster w/ @new2_yj@khflam)
🔶Rapid mixing of stereographic MCMC for heavy-tailed sampling
Fri & Sun (talk by Federica Milinanni w/ @new2_yj)
This week, the @OxfordYss is attending the 2026 World Meeting of the @ISBA_events. An impressive number of our members are in Japan presenting their work.
If you are in Japan this week and would like to discuss their research, please reach out and connect! 🤝
How to incorporate Metropolis-Hastings into score-based Diffusion models to improve generation?
On the coming Tuesday, June 16th 4pm-5pm UK time, we will have @khflam to talk about "Metropolis-Adjusted Diffusion Models" (arxiv.org/pdf/2605.09654) 🚀
Join us with Zoom!! Links 👇
We've lost an absolute giant today. RIP Dimitri Bertsekas. His probability and optimization books got me through my masters. Massive loss for the MIT community and the field.
We propose two approaches:
🔵Exact Barker corrector via a two-coin algorithm: the first exactly adjusted corrector for score-based models
🟢Simpson's rule approximation: efficient with order 5/2 accuracy
Consistent FID gains on FFHQ, AFHQv2 & ImageNet-64 without retraining
[2/3]
[📄preprint] Diffusion models 🤝 MCMC !
Diffusion model samplers are biased due to discretisation
💡The fix: Metropolis-type adjustment on corrector steps
❗️Challenge: no access to the density ratio, only the score
🔑Insight: the score (and some maths) is all you need...
[1/3]
That’s a wrap on the #NeurIPS2025 Main Track! 🌴🏁
We presented 8 papers & loved the energy in San Diego. Huge thanks for the discussions! ☀️
But we aren't done. 🚀 Catch 4 more papers at the Workshops this weekend.
#NeurIPS@OxfordStats
At #NeurIPS2025 presenting 2 papers:
Rao-Blackwellised Reparameterisation Gradients with a rising star @kevinhflam, George Deligiannidis, and @yeewhye
Sparse Gaussian Processes: Structured Approximations and Power-EP Revisited with the inimitable Michalis Titsias
This explains why local reparam gradients have lower variance, but the local reparam trick is only applicable to independent pre-activations.
To lower gradient variance in more general cases, we use conjugate gradients to invert covariance matrices in backprop: yielding R2-G2!
Excited to share our #NeurIPS2025 paper: Rao-Blackwellised Reparameterisation Gradients!
We propose R2-G2 as a general-purpose gradient estimator for latent Gaussians and as the Rao-Blackwellisation of reparam gradients.
Joint work with @thdbui@GeorgeDeligian9@yeewhye
Wrote up some notes providing an introduction to discrete diffusion models, going into the theory of time-inhomogeneous CTMCs via generators/time-evolution systems.
What motivated me was the sheer difficulty of finding a useful reference which laid out the theory (e.g. Kolmogorov equations etc.) in a single place. Hope these can be helpful to others.
927 Followers 4K FollowingLLM Post-Training/ Recursive Self-Improvement for Reasoning
Co-founder/ MTS @ReasonCoreAI
PhD @UCLA, Bachelors @iitbombay
Led a team of ~500 PhD researchers
270 Followers 8K FollowingI am dumb but I am learning.
Curator of AI Curiosities @ https://t.co/EfY6W9zvh1
I email 700K+ ppl about AI every morning. You should follow me.
131 Followers 208 FollowingDPhil in Machine Learning at @UniofOxford advised by @yaringal and @jmhernandez233. Interested in Probabilistic ML and Generative Models.
26K Followers 1K FollowingFind me @[email protected] Professor at @OxCSML, @oxfordstats and Research Director at @GoogleDeepMind. All opinions are my own.
7K Followers 3K FollowingAss. prof. of Generative AI. PI of Generative Memory Lab (@DondersInst). Generative diffusion, language diffusion and statistical physics. AI realist.
1K Followers 58 FollowingNeurIPS Europe is a official satellite event of @NeurIPSConf, that will be held in Paris from Dec 9th to 13th.
To become sponsor: [email protected]
30K Followers 1K FollowingResearch at @GoogleDeepMind. Gemini Omni Team. Priors: GNNs, Structured World Models, Neural Assets, Veo Ingredients/References, Veo Robotics
8K Followers 2K FollowingPrincipal research manager at Microsoft Research Amsterdam. Formerly at Google Brain and University of Amsterdam. PhD in condensed matter physics.
240 Followers 7 FollowingReading group focusing on probabilistic machine learning and molecular applications. Led by @TonyRKOuYang, @JiajunHe614, and @WillLin1028
131 Followers 208 FollowingDPhil in Machine Learning at @UniofOxford advised by @yaringal and @jmhernandez233. Interested in Probabilistic ML and Generative Models.
27K Followers 577 Followingphysics of language models @ Meta (FAIR at MSL, not GenAI or TBD)
🎓:Tsinghua Physics — MIT CSAIL — Princeton/IAS
🏅:IOI x 2 — ICPC — USACO — Codejam — math MCM
701 Followers 437 FollowingAssociate Professor at Georgia Tech. Principal Research Scientist at NVIDIA Research. Areas: AI, Robotics, AI for engineering, Mechanical Engineering, Control
12K Followers 931 FollowingResearch Director, leading fundamental generative AI research (GenAIR) @nvidia research, occasionally volunteer at California Search & Rescue, views are my own.
7K Followers 3 FollowingNew Statistics (Machine Learning) papers from https://t.co/BSsM6dWTu7: supervised, unsupervised. Thank you to arXiv for use of its open access interoperability.