跳到正文
返回论文列表
cs.AI提交于 已译

并非所有证据都可采纳:工具增强医疗智能体的证据治理

Not All Evidence Is Admissible: Evidence Governance for Tool-Augmented Medical Agents

Runxun Zhang · Onofrio Antonio Catalano · Jinqiao Wang · Jingwei Wei

中文摘要

工具增强的医疗智能体将工具返回的任何内容都当作证据。这种默认做法代价高昂:医疗工具会静默失败,而冻结的推理器无法忽略进入其上下文的内容,因此许多原本能正确回答的病例在加入工具输出后被翻转为错误。该现象称为证据劫持(evidence hijacking)。V-信息(V-information)解释了这一现象:推理器失去了忽略的选项,因此工具输出可能携带负的可用信息,而将保留哪些内容的决策移出推理器可恢复该选项。理论进一步表明,任何过滤收益中有一部分仅来自让推理器看到更少内容,只有基于内容的选择才能让它看得更好。据此,将工具增强推理重构为证据治理(evidence governance),把决定什么算作证据与负责回答的推理器分离,并提出上下文无关证据治理(context-independent evidence governance, CIEG),这是首个在获取后对每个工具输出是否可作为证据进入做出裁决的医疗智能体框架。每个工具输出写成可与原始输入核对的卡片;一个仅看到输入和卡片的独立检查器决定哪些卡片成为证据;推理器仅根据被采纳的卡片作答,全程无需训练。在医疗问答、眼底青光眼筛查和胸部X光肺炎分类上,它在同等规模模型中取得最高准确率,并以更低成本显著优于基于相同骨干的最先进工具使用智能体和多智能体系统。

关键要点

  1. 01医疗工具会静默失败,冻结推理器无法忽略工具输出,导致原本正确的答案被翻转为错误,称为证据劫持。
  2. 02V-信息理论表明过滤收益部分仅来自减少可见内容,只有基于内容的选择才能提升推理质量。
  3. 03CIEG框架将证据裁决与推理分离,由独立检查器根据原始输入和卡片决定哪些工具输出可作为证据。
  4. 04在医疗问答、青光眼筛查和肺炎分类上,CIEG在同等规模模型中准确率最高,且成本更低。
  5. 05CIEG全程无需训练,显著优于相同骨干的最先进工具使用智能体和多智能体系统。

解读

尚无解读。

原始英文摘要

arXiv:2611.22169v1 Announce Type: cross Abstract: Tool-augmented medical agents treat whatever their tools return as evidence. We show that this default is costly: medical tools fail silently, and a frozen reasoner cannot ignore what enters its context, so many cases it would have answered correctly are flipped to wrong once tool outputs are added. We call this evidence hijacking. $\mathcal{V}$-information explains the phenomenon: the reasoner has lost the option to ignore, so a tool output can carry negative usable information, and moving the decision of what to keep outside the reasoner restores that option. The theory further shows that part of any filtering gain comes merely from letting the reasoner see less, and only selection by content lets it see better. Building on this, we recast tool-augmented reasoning as evidence governance, which separates deciding what counts as evidence from the reasoner that answers, and propose context-independent evidence governance (CIEG), the first medical agent framework that rules, after acquisition, on whether each individual tool output may enter as evidence. Every tool output is written as a card that can be checked against the raw input; an independent checker that sees only the input and the cards decides which cards become evidence; and the reasoner answers from the admitted cards alone, with no training anywhere. On medical question answering, fundus glaucoma screening, and chest X-ray pneumonia classification, it achieves the highest accuracy among models of comparable scale, and significantly outperforms state-of-the-art tool-using agents and multi-agent systems built on the same backbone at lower cost.

同方向论文 · cs.AI

查看全部 →

反向发现 · 这篇论文相关的工具与提示词

去交叉发现 →

关联提示词 · 1

研究github.com/CHENJIAMIAN/awesome-chatgpt-prompts-zh-CN·外部收录·

LLM 研究员

我要你充当大型语言模型(Large Language Model)研究领域的专家。请仔细阅读用户提供的论文、文本或概念性术语,然后回答他们提出的问题。在作答时,请确保不遗漏任何重要细节。基于你的理解,你还应说明该概念背后的原理、实现过程及目的。如有可能,你可利用网络搜索查找有关该概念或其推理过程的额外信息。在呈现信息时,凡有可用的论文引用或链接,均须一并提供。