Pinned
My first blog post in over a year is a deep dive on flow maps🗺️, or how to learn the integral of a diffusion model to enable faster sampling and several other cool tricks.
It's the longest one yet👀 Let me know what you think!
Sander Dieleman
@sedielem
Research Scientist at Google DeepMind (WaveNet, Nano Banana, Gemini Omni). I tweet about ML, music, generative models (personal account).
- Not one, but two diffusion LM workshops at NeurIPS this year! Both in Sydney. If they're both on the same day as well, that's going to make for some interesting scheduling challenges, looking at the invited speaker lists😂 7amin.github.io/diffulm-neurip… bento-neurips.github.io/#speakersWe are happy to announce the list of accepted NeurIPS 2026 workshops: blog.neurips.cc/2026/08/10/ann… The workshops will take place on: - Fri Dec 11 and Sat Dec 12, 2026 – Sydney - Sat Dec 12 and Sun Dec 13, 2026 – Paris and Atlanta We want to thank everyone who submitted a proposal
- Nice blog post about the current state of discrete diffusion language models. Although it explicitly does not cover continuous methods, it provides useful context for understanding how those might change the associated trade-offs. Stay tuned for more on that from me as well!👀Some Theoretical and Practical Thoughts on Diffusion Language Models jzhao2024.github.io/notes/2026/08/…
- Excellent talk by @thjashin: - What separates autoregression and diffusion? - Bridging the gap between continuous and masked discrete diffusion via loss reweighting - Insertion-based sequence generation Very clear and easy to follow! (via @mblondel_ml)
- The Million Song Dataset challenge! kaggle.com/c/msdchallenge I did a deep dive into collaborative filtering for this, and ended up using weighted matrix factorisation, finishing 21st. A much simpler and intuitive method took 1st place. A valuable lesson in my early PhD years🤓
