Reproduce “Program-Based Strategy Induction for Reinforcement Learning” 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
Carlos G. Correa, Thomas L. Griffiths, Nathaniel D. Daw
The authors search for learning strategies as small programs with memory-update and choice rules, using Bayesian inference to balance expected reward against a simplicity-favoring prior. In a two-arm bandit, an accumulating-memory strategy with stochastic choice reproduces the direction of horizon-dependent randomness in equal-information conditions, while becoming more decisive over time. This is a model account, not evidence people use programs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.