Multi-Agent Debate

Researchers propose Debate-on-Graph (DoG), a framework that enables large language models (LLMs) and uncertain knowledge graphs (UKGs) to collaborate adaptively for reliable question answering. DoG us…