Reproduce “Aristotle: Mastering Logical Reasoning with A Logic-Complete Decompose-Search-Resolve Framework” 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

Jundong Xu, Hao Fei, Meng Luo, Qian Liu, Liangming Pan, William Yang Wang, Preslav Nakov, Mong-Li Lee and 1 more

Aristotle, a logic-complete framework, integrates symbolic logic into large language models' decompose-search-resolve processes to enhance logical reasoning. It consistently outperforms state-of-the-art baselines by up to 11.6% in accuracy while significantly improving computational efficiency across various complex logical tasks.

Key insights

  • Deeply integrating formal logical principles (like CNF, proof-by-contradiction, and resolution) into an LLM's reasoning modules dramatically improves both the accuracy and efficiency of logical inference.
  • A rule-based, logic-driven search mechanism is more reliable and efficient for navigating logical search spaces than heuristic or LLM self-evaluation strategies.
  • LLMs, when guided by a structured, logic-complete framework, can effectively perform high-level symbolic logical reasoning internally, demonstrating a path towards more robust neurosymbolic AI.

Summary by alphaXiv.

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