Mining Math Conjectures from LLMs: A Pruning Approach
This work addresses the challenge of automating mathematical discovery for researchers in group theory, though it is incremental as the conjectures are not groundbreaking.
The researchers tackled the problem of generating mathematical conjectures in group theory using Large Language Models, demonstrating that LLMs can produce original conjectures about the solubilizer construct, with results showing they generate plausible or falsifiable conjectures but have limitations in code execution.
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.