Reproduce “Backbone-Informed Adaptation after EEG Pretraining: A Capacity-Controlled Study” 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

Mingyang Jiang, Yamin Li, Fan Ma, Daniel Moyer, Hua Xu, Catie Chang

The authors test whether channel- and time-aware adapters improve frozen EEG features over parameter-matched token-wise MLPs. The native adapter mixes along channel and temporal-patch axes; the control transforms each feature vector independently. Across six primary comparisons, there is no consistent native-adapter advantage: on LaBraM–FACED, macro-F1 is 0.1406 ± 0.0049 versus 0.1940 ± 0.0024, and results vary by task and protocol.

Summary by alphaXiv.

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