Multiagent Debate

Multi-agent debate framework research project that improves reasoning accuracy by having multiple LLM instances debate each other. Experiments demonstrate that the debate mechanism significantly enhances correctness in mathematical and logical reasoning.