Meta's TRIBE v2: Tri-modal foundation model predicts human brain activity from 1,000+ hours of fMRI
burny_tech · x · 2026-09-28
Stéphane d'Ascoli, Jean-Rémi King and colleagues at Meta introduced TRIBE v2, a tri-modal (video, audio, language) foundation model that predicts human brain activity across naturalistic and experimental conditions.
- Unified training on 1,000+ hours of fMRI from 720 subjects
- Several-fold accuracy improvements over traditional linear encoding models on novel stimuli, tasks and subjects
- Enables in silico experimentation: reproduces decades of established results in classic visual and neuro-linguistic paradigms
- Interpretable latent features reveal fine-grained topography of multisensory integration
Paper: arXiv:2605.04326.
More from AGI Musings
- Effective Altruism is top-down and hierarchical, not a grassroots movement — ctjlewis · 2026-11-16
- Dev debunks 'rogue AI' hype: subsidized subscriptions can't fund expensive swarms — gerardsans · 2026-11-16
- Ben Thompson: The Demise of Jensen Huang Marks the End of Nvidia's Prospects and the AI Bubble — morqon · 2026-11-16
- Blogger: 5-10 More GW of Compute and Tens of Millions of AI Scientists Could Reach 'Magic' — Dr_Singularity · 2026-11-16
- Five case studies unpack how Google AI Overviews hit clicks and CTR in France — gaganghotra_ · 2026-11-16
- lemire: AI's real impact on coding is expanding the engineer's 'cognitive radius', not speed — lemire · 2026-11-16