Reproduce “CogRSI: Evolving and Auditing Cognitive Reasoning Strategies” 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

Yandan zheng, Haoran Luo, Xinliang Zhou, Anhtuan Luu

CogRSI marks who witnessed each story sentence, retains the full story, and compares annotated with plain readings. On the eight-benchmark test sample, Qwen3.5-35B-A3B scores 73.1% Main versus 68.7% for nine-sample self-consistency at matched token cost. Its learned gated library reaches 96.6% pair validity on BigToM-U, but most decided rules audited were invalid.

Summary by alphaXiv.

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