发表机构
City University of Hong Kong; Fudan University Shanghai Cancer Center; Affiliated Hangzhou First People’s Hospital, Westlake University School of Medicine; Zhejiang University; Shaoxing Vocational & Technical College（香港城市大学; 复旦大学附属肿瘤医院; 西湖大学医学院附属杭州市第一人民医院; 浙江大学; 绍兴职业技术学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究构建肿瘤决策边界基准评估9个前沿LLMs，发现其存在临床元判断盲点，需架构干预，模型质量非临床部署主瓶颈，需可检测自身能力边界并路由决策至临床医生的架构。
AI 中文摘要
大语言模型（LLMs）在医学知识考试中取得高分，但现实世界的肿瘤学并非知识测试——它是一系列指南路径选择、升级判断和不确定性下的承诺。现有基准主要衡量事实回忆，而前沿LLMs是否存在无法通过组合模型解决的决策路径盲点仍不明确。我们构建了肿瘤决策边界基准（ODBB），涵盖NCCN指南和结直肠癌病例中的2005个肿瘤决策点，并评估了2025年6月至2026年4月发布的9个前沿LLMs（4个闭源、5个开源权重系列）。一个完全确定性评分器（零LLM推理）将输出分为14种失败类型，由两名肿瘤医生对225项分层样本独立验证（Cohen加权κ分别为0.939和0.790）。将9个模型视为合并超级模型，42.1%（Wilson 95%置信区间40.0--44.3%）的所有项目——1586个NCCN项目中的35.7%和419个结直肠癌病例中的66.4%——未被任何模型答对，失败集中在在任何推理前选择指南路径之间：这是临床元判断中一致的盲点，可能需要架构干预而非更多训练数据。两个针对果断性调优的模型（GPT-5.5、Gemini 3.1 Pro Preview）的不安全承诺频率是7个谨慎模型的3至5倍，且得分未更高。在3--9%的项目中，模型陈述了正确的下一步临床步骤但未作出承诺——这是决策失败，而非知识失败。模型质量不再是临床LLM部署的主要瓶颈；关键约束是认为任何单一模型可作为临床决策唯一基础的假设。进展需要能检测模型达到能力边界并将决策路由给临床医生的架构。
英文摘要
Large language models (LLMs) achieve high scores on medical knowledge examinations, yet real-world oncology is not a knowledge test--it is a sequence of guideline-pathway choices, escalation judgments, and commitments under uncertainty. Existing benchmarks largely measure factual recall, leaving open whether frontier LLMs share decision-path blind spots that combining models cannot fix. We built the Oncology Decision Boundary Benchmark (ODBB)--2,005 oncology decision points across NCCN guidelines and colorectal cancer cases--and evaluated nine frontier LLMs (four closed-source, five open-weight families) released between June 2025 and April 2026. A fully deterministic scorer (zero LLM inference) classified outputs into 14 failure types, independently validated by two oncologists (Cohen's weighted $κ$ = 0.939 and 0.790) on a 225-item stratified sample. Treating the nine as a pooled super-model, 42.1% (Wilson 95% CI 40.0--44.3%) of all items--35.7% of the 1,586 NCCN items and 66.4% of the 419 colorectal-cancer cases--were answered correctly by none, with failures concentrated in choosing between guideline pathways before reasoning within any: a consistent blind spot in clinical meta-judgment that likely requires architectural intervention rather than more training data. Two models tuned for decisiveness (GPT-5.5, Gemini 3.1 Pro Preview) made unsafe commitments three to five times more often than the seven cautious models without scoring higher. In 3--9% of items, models stated the correct next clinical step yet did not commit to it--failures of decision, not knowledge. Model quality is no longer the primary bottleneck for clinical LLM deployment; the binding constraint is the assumption that any single model can be the sole basis for a clinical decision. Progress requires architectures that detect when a model reaches its competence boundary and route the decision to a clinician.
