Multi-Agent Debate

Several agents independently reason and then critique each other's answers over rounds to converge on a more factual, better-verified result.

grounded in: Established multi-agent reasoning technique in this system's domain; extends the existing multi-agent / llm-as-judge / verification gate concepts as a hallucination-reduction method.

Connected concepts

Multi-agent systems, LLM-as-judge, Self-Consistency, Verification loops, Hallucination control