RMA — agentic framework tackles research-level math problems
arXiv paper introduces Research Math Agents (RMA), a multi-agent system built for long-horizon mathematical reasoning at research scale—not competition math or formal proofs.
• Specialized modules: problem analysis, literature search, fair comparison, knowledge-bank construction, proof verification
• Initializer, proposer, and verifier agents coordinate via shared structured memory
• Multi-role, multi-round workflow with iterative proof refinement