---
id: 20260404-T0-09
title: "临床预测AI采用多代理 deliberation 处理复杂病例"
title_en: "Multi-Agent Deliberation Boosts Clinical Prediction Accuracy"
url: https://ai.daily.yangsir.net/daily/20260404-T0-09
issue_date: 2026-04-04
publish_date: 2026-04-03T04:00:00.000Z
category: research
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2604.00085
---

# 临床预测AI采用多代理 deliberation 处理复杂病例

大模型用于临床预测时，简单病例结果一致，复杂病例因提示变化差异显著。研究提出案例自适应多代理deliberation框架，通过多个专业代理协作处理复杂病例。实验显示，在心脏病和糖尿病预测任务中，该框架准确率提升15%，特别适用于罕见病和并发症判断。医院可用此框架构建更可靠的AI诊断辅助系统，减少医生误诊风险。

## English Version

**Multi-Agent Deliberation Boosts Clinical Prediction Accuracy**

LLMs show inconsistent results for complex clinical cases. The case-adaptive multi-agent deliberation framework uses specialized agents collaborating to handle difficult cases. It improves prediction accuracy by 15% for heart disease and diabetes, particularly effective for rare diseases and complications. Hospitals can deploy this for more reliable AI diagnostic assistance.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2604.00085)

**详情页**：https://ai.daily.yangsir.net/daily/20260404-T0-09

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*