Researchers have developed RABBiT, a novel audio-to-fMRI encoder designed to predict brain responses to speech with high accuracy in zero-shot and few-shot scenarios. This model significantly outperforms existing state-of-the-art methods and group averages by incorporating learned region-specific attention and decomposing brain responses into shared and subject-specific components. RABBiT's parameter-efficient tuning allows for substantial performance improvements with minimal participant-specific data, enabling more scalable analyses of language in the human brain. AI
IMPACT Enables more scalable population-level analyses of language in the human brain by improving prediction accuracy and reducing data requirements.
RANK_REASON The cluster contains an arXiv preprint detailing a new research model and methodology.
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- functional magnetic resonance imaging
- Gotit.pub
- Hugging Face
- RABBiT
- ScienceCast
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
