Reproduce “A small cognitive core explains human behavior at scale” 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
Marcel Binz, Eric Schulz
CORE jointly models behavior across psychological experiments using layered, context-sensitive memory that stores prediction errors, retrieves experiences, and forgets them. It represents experiment histories in natural language and predicts choices. Across 137 experiments, CORE had lower held-out loss than a matched Transformer in all 62 studies and lower overall loss than Centaur in 47 of 62 studies; it remains an initial framework, not a complete theory.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.