Diffusion transformers generate fMRI time series: NeurIPS 2026 paper explores in-silico neuroscience
LucaAmb · x · 2026-11-06
A NeurIPS 2026-accepted preprint trains diffusion transformers to generate neural (fMRI) time series, conditioning per timestep by injecting conditioning tokens directly into context, evaluated across hundreds of unseen task conditions. The authors will also present a flow-models-for-scientific-interactions paper as an oral at the AIDaR workshop. Preprint: arxiv.org/abs/2610.20288.
More from Research
- Mech interp researchers push back: 'not causal' shouldn't end every argument — burny_tech · 2026-11-06
- 'Say more than not causal': descriptive and causal mech interp work inform each other — burny_tech · 2026-11-06
- UChicago HCI lab releases UIST 2026 slides: RingIO and electrotactile thermal referral — plopesresearch · 2026-11-06
- EnvOpt Paper: Optimizing Agent Environments Cuts Cost 72.7% and Boosts Pass Rate 14.34% — dair_ai · 2026-11-06
- Team beats Mem0 on its own benchmark after 6 weeks, then hits a 7x-too-big submission bug — gauzzastrip · 2026-11-06
- Reviewer scandal: Pangram flags 3 of 4 peer reviews as 100% AI-generated on a rejected paper — nsaphra · 2026-11-06