SCRIBE Framework Bridges LLM Semantics with Human Brain Activity
October 16, 2026
SCRIBE uses an ensemble of open-source LLMs to create word-level semantic representations that reach 75-81% of the human noise ceiling. The framework successfully predicts neural responses in language-selective brain regions during naturalistic listening tasks.
HOW THIS AFFECTS YOU
●
researcherYou can use this to study how contextual linguistic meaning is represented in the human brain using scalable, LLM-derived features.