Tag
该论文提出 PsyCIDRA，一个将自由形式精神科访谈与诊断推理相结合的双智能体框架，其中访谈智能体借助工具维护笔记、加载专家技能并检索 ICD-11 参考，诊断智能体则基于访谈记录生成假设并报告支持、冲突和缺失证据。模拟评估和针对 101 名人类参与者的盲测研究表明，该框架的诊断一致性显著优于直接提示。
Introduces LingxiDiagBench, a large-scale multi-agent benchmark for evaluating LLMs on Chinese psychiatric consultation and diagnosis. Key findings show high accuracy on binary classification but poor performance on multi-way differential diagnosis, highlighting a decoupling between conversational quality and diagnostic accuracy.
WiseMind is a knowledge-guided multi-agent framework that uses LLMs for psychiatric diagnosis by combining a "Reasonable Mind" agent for evidence-based logic with an "Emotional Mind" agent for empathetic communication, achieving 85.6% diagnostic accuracy on simulated and real patient interactions. The framework leverages DSM-5 structured knowledge graphs to reduce hallucinations and outperforms single-agent baselines by 15-54 percentage points while maintaining clinical soundness and psychological support.