Reproduce “Mixed Signals: Weight-Space Programming for EEG and ECG” 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
Dara Varam, Taraneh Aminosharieh Najafi, David Atienza
The authors demonstrate weight-space programming by having Claude Opus 5.5 write all 2,730 weights of a compact EEG classifier for eyes-open versus eyes-closed detection, without gradient descent. Given the task, network design, and recordings from five practice subjects, it reached 0.834 balanced accuracy on 65 unseen subjects, matching training on 30 subjects; this does not establish reliable weight-reading.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.