arXiv Artificial Intelligence· Zihao Zheng (Washington University in St. Louis), Junyi Yao (Washington University in St. Louis), Baichuan Li (Southern Methodist University), Jiayu Long (Washington University in St. Louis)·· 2 小时前AI 评分29
医疗 LLM 智能体升级可靠性审计：工作流状态失效的扰动控制基准
Auditing Escalation Reliability in Healthcare LLM Agents: A Perturbation-Controlled Benchmark of Workflow-State Failures
AI 导读
研究构建了 400 个合成医疗工作流 episode 的评估协议，在 5:1 漏报与不必要升级成本下考察 ANSWER/ASK CLARIFICATION/ABSTAIN/ESCALATE 决策。
来源：arXiv Artificial Intelligence · arxiv.org