Cajal-1: 218M-param whole-cortex encoding model predicts fMRI from any video or audio
AfterShock182 · reddit · 2026-10-19
The author released Cajal-1, a 218-million-parameter whole-cortex encoding model covering vision, audition, and language, trained on fMRI data from 2,000 people watching films, TV, and images.
On the films Meta held out to test TRIBE v2, Cajal-1 is 25% more accurate and closer to the real brain in 27 of 32 recordings. Given any video, podcast, or song, it predicts fMRI responses at 20,484 points across the human cortex every 1.49 seconds — no scanner needed. The model and a Colab demo are open on Hugging Face. It's named after Santiago Ramón y Cajal, who showed in the 1890s that the brain is built from separate cells.
More from Research
- EMNLP paper: diffusion LMs internally encode latent denoising timestep representations — gsarti_ · 2026-10-23
- KU Leuven and VIB open two Group Leader professorships in AI for Biology, start 2027 — jpillowtime · 2026-10-23
- Subspace Tuning curbs LLM catastrophic forgetting by freezing principal singular directions — scottniekum · 2026-10-23
- Perceptic launches PercepticOS, an AI workspace for 'scientific superintelligence' — nathanbenaich · 2026-10-23
- EPFL's Bosselut to talk long context as test-time learning at ELLIS seminar — andre_t_martins · 2026-10-23
- DRATS: Adaptive Task Sampling Boosts Worst-Task Performance in Multi-Task RL — Cohere · 2026-10-23