Reproduce “MoodAngels: A Retrieval-augmented Multi-agent Framework for Psychiatry Diagnosis” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mengxi Xiao, Ben Liu, He Li, Jimin Huang, Qianqian Xie, Xiaofen Zong, Mang Ye, Min Peng
Researchers from Wuhan University developed MoodAngels, a retrieval-augmented multi-agent framework for mood disorder diagnosis that incorporates granular-scale analysis and a multi-agent debate mechanism. This framework achieved an accuracy of 0.925 on real clinical data, outperforming bare GPT-4o by 12.3%, and is supported by MoodSyn, an open-source synthetic psychiatric dataset with 1,173 cases.
Key insights
- Granular-scale analysis of individual assessment scale questions, rather than total scores, and consistency evaluation across symptom groups is crucial for resolving diagnostic discrepancies and improving accuracy in psychiatric AI.
- Integrating a retrieval datastore with structured DSM-5 criteria and anonymized historical clinical cases significantly enhances the AI's ability to handle symptom overlap and complex diagnostic scenarios.
- A multi-agent debate framework yields more robust, clinically coherent, and interpretable diagnoses compared to single-agent or bare LLM approaches, especially in ambiguous cases.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.