Reproduce “Decomposing Brain-LLM Alignment with Interpretable Semantic Representations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Camila Kolling, Mariya Toneva
The authors compare brain responses from seven listeners hearing Moth Radio Hour stories with LLaMA2-7B and SCRIBE, a representation of contextual semantic properties. They partition held-out fMRI prediction into information shared by both representations and information unique to each. For layer 15, the shared component explains 0.028 whole-brain variance, but findings are limited to this model, layer, dataset, and sample.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.