Reproduce “Interpretable weight-space programming builds transparent transformers and brain encoding models” 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

Yash Mehta, Chandan Singh, Weijia Xu, Richard J. Antonello, Adeen Flinker, Nima Mesgarani, Michael F. Bonner, Nebojsa Jojic and 1 more

iWSP uses an LLM to write interpretable brain-encoding features as code and revise them against a fixed evaluator, without training the model. Audits and description-based rebuilds assess whether the resulting models remain understandable. In audio ECoG, combining iWSP and Whisper features raised electrode-wise Pearson correlation from 0.070 to 0.075 on one patient’s hidden session, though 40 of 166 electrodes worsened.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.