Reproduce “Generating Computational Cognitive Models using Large Language Models” 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

Milena Rmus, Akshay K. Jagadish, Marvin Mathony, Tobias Ludwig, Eric Schulz

The GeCCo pipeline leverages large language models (LLMs) to automate the generation and iterative refinement of computational cognitive models. It produced interpretable models that consistently matched or surpassed hand-crafted baselines across four distinct cognitive domains, capturing explainable variance in human behavior comparable to a cognitive science foundation model.

Key insights

  • Large Language Models can effectively generate interpretable computational cognitive models that are competitive with or superior to expert-designed baselines.
  • The iterative feedback mechanism within GeCCo is the most critical component for driving the improvement and refinement of generated models.
  • LLM-generated models can provide novel theoretical insights, such as single-parameter arbitration between heuristics or distinct learning mechanisms, challenging established cognitive theories.

Summary by alphaXiv.

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