arXiv cs.AI· Yunqi Zhu, Wen Tang, Jinghao Niu, Wensheng Zhang, Xuebing Yang·· 2 天前AI 评分27
MedTraj：面向医疗智能体的临床推理轨迹构建与评估框架
Constructing and Evaluating Clinical Reasoning Trajectories for Medical Agent
AI 导读
研究提出 MedTraj 框架，将医疗 AI 智能体的推理轨迹作为构建、评估与优化对象，把多步临床推理链解析为临床观察、证据、编号推理步骤与结论，并从连贯性、证据支持、幻觉、完整性、可追溯性五个维度打分。
正文 · 原文
Abstract:Evaluation of medical artificial intelligence agents remains predominantly answer-centric, assessing only the correctness of final outputs while overlooking the quality of intermediate reasoning. In clinical settings, however, a correct answer reached through fabricated evidence or incoherent logic is as dangerous as an incorrect one. We propose MedTraj, a framework that treats reasoning trajectories as critical objects for construction, evaluation, and optimization. The pipeline generates structured multi-step clinical reasoning chains from medical diagnosis sources. Each trajectory is then parsed into clinical observations, evidence, numbered reasoning steps, and a final conclusion, and scored across five quality dimensions: coherence, evidence support, hallucination, completeness, and traceability. Controlled error injection introduces targeted faults into otherwise correct trajectories to establish causal links between specific reasoning failures and measurable quality degradation. Building on this, step-level filtering based on marginal contribution identifies which individual reasoning steps drive or undermine trajectory quality. Finally, quality-weighted context learning feeds trajectory evaluations back into the model at inference time, allowing it to learn from both strong and weak reasoning demonstrations. Experiments across CareQA, PubMedQA, and CECMed demonstrate that trajectory context consistently improves reasoning coherence, with gains of +0.029 to +0.041 over a zero-shot baseline. On the CECMed severity assessment task, quality-weighted context nearly doubles the correctness over the zero-shot baseline while cutting the hallucination ratio by 87%. Marginal-contribution analysis further shows that a small minority (15.3%) of reasoning steps carry most of the quality signal, and that extending chains beyond four steps yields diminishing returns.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.05090 [cs.AI] |
| (or arXiv:2609.05090v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.05090 arXiv-issued DOI via DataCite |
Submission history
From: Yunqi Zhu [view email]
[v1]
Fri, 4 Sep 2026 12:44:10 UTC (2,773 KB)
[v2]
Thu, 15 Oct 2026 06:31:38 UTC (8,404 KB)
来源：arXiv cs.AI · arxiv.org