Reproduce “Synthesizing symbolic models of human and machine reasoning 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
Solim LeGris, Philipp Mondorf, Jinran Jin, Akshay K. Jagadish, Brenden M. Lake, Todd M. Gureckis
The authors introduce BCSS, a method for interpreting incorrect answers on colored-grid reasoning tasks by proposing edits to a correct symbolic program. An LLM suggests constrained edits, which are executed and retained only when they reproduce an observed response. BCSS reproduced 3,636 of 7,402 filtered human errors (49.1%), yielding candidate rules, not verified accounts of participants’ reasoning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.