发表机构
Stanford University; Emory University; Mayo Clinic（斯坦福大学; 埃默里大学; 梅奥诊所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对LLM决策存在的不忠实、不一致问题，提出VERDICT智能体，将决策任务转化为SMT问题推导决策，在临床试验匹配任务中实现高准确性、完美政策一致性及受临床医生青睐的可问责理由。
AI 中文摘要
可问责性指的是一项决策能够被审查、证明合理并提出质疑。大型语言模型（LLM）使得这一过程变得困难：其流畅的输出可能缺乏依据、不完整，或与决策过程不忠实。实现可问责性需要经过验证的理由（决策是如何做出的）、假设（被假定而非已知的内容）、政策一致性（对相同事实采用相同处理方式）以及关键条件（哪些因素会改变结果）。我们将自我忠实性作为可问责性的自动测试：改变关键条件应改变决策。我们通过临床试验匹配来研究可问责人工智能，这是循证医学中的一项高风险核心任务。尽管基于LLM的匹配器能以合理的准确性将患者与试验匹配，但它们应用决策政策时不一致，且生成的理由与自身决策不忠实。我们推出VERDICT，这是一种基于LLM的智能体，它将决策任务、其约束条件及政策转化为模理论可满足性问题（SMT），随后使用SMT和MaxSMT求解器推导决策——因此政策应用一致，且决策从结构上具备可问责性。在源自SIGIR 2016的数据集和TREC 2021数据集上，VERDICT在纯LLM和神经符号基线中实现了最强的决策准确性，政策应用达到完美一致性，且生成的理由以明确的假设和关键条件为依据，更受临床医生青睐，同时反事实自我忠实性也有所提升。
英文摘要
Accountability means a decision can be examined, justified, and contested. LLMs make this hard: fluent output may be ungrounded, incomplete, or unfaithful to the decision process. Achieving accountability requires verified rationales (how was the decision reached), assumptions (what was assumed rather than known), policy consistency (the same treatment for the same facts), and pivotal conditions (what would change the outcome). We introduce self-faithfulness as an automatic test of accountability: changing the pivotal conditions should change the decision. We examine accountable AI through clinical trial matching, a high-stakes task central to evidence-based medicine. Although LLM-based matchers match patients to trials reasonably accurately, they apply decision policies inconsistently and produce rationales that are unfaithful to their own decisions. We introduce VERDICT, an LLM-based agent that translates a decision task, its constraints, and its policy into Satisfiability Modulo Theories (SMT), then derives the decision with SMT and MaxSMT solvers -- so policies are applied consistently and decisions are accountable by construction. Across a SIGIR 2016-derived dataset and TREC 2021, VERDICT achieves the strongest decision accuracy among LLM-only and neurosymbolic baselines, applies policies with perfect consistency, and produces clinician-preferred rationales grounded in explicit assumptions and pivotal conditions, with improved counterfactual self-faithfulness.
CommentsAccepted to EMNLP 2026 (Main Conference). 46 pages, 6 figures, 28 tables. Code and prompts: https://github.com/stanford-oval/clinical-trial-matching
