Reproduce “Scaling Neural Motor Decoding via Decoupled Behavioral Pretraining” 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

Divyansha Lachi, Vinam Arora, Shivashriganesh P. Mahato, Alexandre Andre, Eva L. Dyer

BEEMO lets neural decoders learn movement patterns from behavior recordings collected without brain data, then uses neural activity to condition movement predictions. It combines this behavior pretraining with paired neural-behavior data, rather than replacing neural recordings. In nonhuman-primate motor tasks, it matched full-data cross-task performance using 32 neural sessions (about 7 hours), versus 105 sessions for full-data backbones.

Summary by alphaXiv.

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