📄 中文摘要
在 AI 辅助研究中，存在一个鲜为人知的危机。现有的工具如 ChatGPT、Perplexity、Copilot 和 Elicit 等，通常只提供一个自信的答案，缺乏对不同观点的全面分析。为了解决这一问题，提出了一种新的 AI 系统，结合了 GraphRAG 和多模型委员会的方式，使四个智能体能够对科学论文进行辩论。通过这种方法，系统能够识别并指出不同智能体之间的分歧，从而提供更可靠的答案。这种多智能体的辩论机制不仅提高了信息的准确性，还增强了研究的透明度和可信度。
我构建了一个让四个智能体辩论科学论文的 AI 系统，并告诉你它们的分歧所在
📄 English Summary
I Built an AI System That Makes 4 Agents Debate Scientific Papers , And Then Tells You Where They Disagree
A quiet crisis exists in AI-assisted research, where tools like ChatGPT, Perplexity, Copilot, and Elicit provide only one confident answer, lacking comprehensive analysis of differing viewpoints. To address this issue, a new AI system is proposed that combines GraphRAG with a multi-LLM council, enabling four agents to debate scientific papers. This approach allows the system to identify and highlight disagreements among the agents, resulting in more trustworthy answers. The multi-agent debate mechanism enhances both the accuracy of information and the transparency and credibility of research.