New IC-based fMRI encoding models predict functional brain networks during stories
neuranna · x · 2026-08-04
A new arXiv paper proposes independent-component-based encoding models for brain activity during story comprehension. Instead of predicting voxels or anatomical regions directly, the method predicts functional networks derived from ICA on held-out fMRI data.
Key points:
- fMRI is decomposed into independent components using one subset of the data.
- Encoding models are then trained on separate data to predict IC time series from LLM representations of the linguistic input.
- Across subjects, a subset of components shows consistently high predictivity.
- These highly predicted components are spatially and temporally consistent and include auditory and language networks.
- Components identified as noise or motion artifacts perform poorly, suggesting the method is capturing genuine stimulus-driven signals rather than confounds.
The paper argues this improves interpretability and reduces problems from voxelwise redundancy and inter-subject variability.
More from Research
- AI paper argues best-of-K boosts generative expressivity, not just sampling quality — anshulkundaje · 2026-08-04
- ASCII art may be a better taste benchmark for frontier models than you think — weswinder · 2026-08-04
- A curated reading list for DeltaNet, FlashKDA, vLLM serving and MoE — austinvhuang · 2026-08-04
- AI index steepens 5x after late 2024 as compute shifts from pretraining to inference — ProfBuehlerMIT · 2026-08-04
- Free app teaches LLM basics and trains a small model locally on Apple MLX — dr_cintas · 2026-08-04
- Pure VLAs may not need long-horizon planning if VLMs can cover it — m_wulfmeier · 2026-08-04