Mining Math Conjectures from LLMs: A Pruning Approach

Jake Chuharski, Elias Rojas Collins, Mark Meringolo

Keywords: Conjecture, Math, Language Models, LLMs, Solubilizer, Abstract Algebra

TL;DR: Presents a successful approach to getting original math conjectures out of language models and then pruning them to get rid of obviously false ones.

Abstract: We present a novel approach to generating mathematical conjectures using Large Language Models (LLMs). Focusing on the solubilizer, a relatively recent construct in group theory, we demonstrate how LLMs such as ChatGPT, Gemini, and Claude can be leveraged to generate conjectures. These conjectures are pruned by allowing the LLMs to generate counterexamples. Our results indicate that LLMs are capable of producing original conjectures that, while not groundbreaking, are either plausible or falsifiable via counterexamples, though they exhibit limitations in code execution.

Concurrent Submissions: N/A

Submission Number: 59