Reproduce “A New Strategy for Artificial Intelligence: Training Foundation Models Directly on Human Brain Data” 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
Maël Donoso
The authors propose a theoretical framework for adding brain measurements to foundation model training as a complement to human-produced text and ratings. It outlines RLHB, which uses valuation-related neural signals alongside preference rewards, and CoTHB, which uses execution-related signals to guide reasoning. The paper synthesizes prior brain-guided learning work but reports no experiments showing either method improves foundation models.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.