RMA: an Agentic System for Research-Level Mathematical Problems
Researchers have developed RMA, an agentic system designed to tackle complex, research-level mathematical problems. This framework breaks down the proof-solving process into specialized modules for analysis, literature review, and verification, coordinated by multiple agents. RMA demonstrated superior performance on the First Proof benchmark, solving eight out of ten problems and generating more robust proofs compared to existing systems like GPT-5.2R and Aletheia. AI
IMPACT This system could accelerate AI's ability to contribute to novel mathematical discovery and formal verification.