Abstract
Recent advances in AI for Mathematics have focused largely on autoformalization and theorem proving, leaving the role of Computer Algebra Systems (CAS) in agentic LLM workflows underexplored. We propose a ReAct-style agentic setup that combines LLM reasoning with verifiable feedback from SageMath, alongside Context7 for up-to-date documentation.
We evaluate this agentic setup across frontier models for solving research-level mathematical problems from the RealMath benchmark in a setting emulating a computational-mathematics research loop.
Additionally, we propose a refinement to the RealMath benchmark by introducing a multi-step post-processing procedure and a multi-stage validation pipeline, both enhancing the quality and reliability of the extracted problem set.
Our experiments reveal substantial performance gains from SageMath access across all evaluated models, averaging a 9.7 percentage point increase, with gains ranging from 1.5 to 27.8 percentage points, narrowing the gap between open-weight and closed models. Qwen 3.7-Max benefits most from SageMath, while GPT-5.5 achieves the highest solve rate of 75.2% and the lowest token usage among tool-enabled configurations.
Our findings suggest that CAS-augmented agents represent a promising direction for assisting mathematicians in computational exploration, and we believe this work is a step towards automated conjecture discovery. The project repository is available online.
Blogger's Review: This study showcases the potential of integrating Computer Algebra Systems with Large Language Models, particularly their effectiveness in solving complex mathematical problems. We look forward to seeing more interdisciplinary collaborations like this to advance the automation of mathematical research.