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 lets a coding agent write interpretable model features directly and revise them using a fixed evaluator’s score, rather than learning feature weights through backpropagation. In brain encoding, the evaluator fits ridge-regression readouts. The visual model outperforms interpretable baselines on held-out data in early and ventral cortex, but trails AlexNet in ventral cortex; story-position features also reveal scan-edge timing.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.