What it measures

Clinical language models face seventeen datasets spanning diagnosis, recommendations, summarization, patient education, and information extraction. Published results keep each task and metric separate.

Reading the measurement

Accuracy is only part of the picture. Inspect unsupported answers, abstentions, language coverage and the evidence available to the model. A confident response can still be wrong, and a broad average can hide a specialist gap.

Published evaluation

Source-reported zero-shot score on the named ClinicBench task. Accuracy, ROUGE-L, entity F1 and micro F1 remain distinct measures; no composite score is invented across them. This paper’s ClinicBench is separate from NEJM Case Records and from later projects with the same name.

Where this fits in an evaluation suite

Read accuracy together with error and abstention rates. Domain averages can hide language gaps, specialist failures, and poorly calibrated confidence.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Source-reported zero-shot score on the named ClinicBench task. Accuracy, ROUGE-L, entity F1 and micro F1 remain distinct measures; no composite score is invented across them. This paper’s ClinicBench is separate from NEJM Case Records and from later projects with the same name.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. Large Language Models in the Clinic: A Comprehensive Benchmark — published result table ↗

    Liu, Fenglin, Li, Zheng, Zhou, Hongjian, Yin, Qingyu, Yang, Jingfeng, Tang, Xianfeng et al. · 2024

    All 19 authors

    Liu, Fenglin, Li, Zheng, Zhou, Hongjian, Yin, Qingyu, Yang, Jingfeng, Tang, Xianfeng, Luo, Chen, Zeng, Ming, Jiang, Haoming, Gao, Yifan, Nigam, Priyanka, Nag, Sreyashi, Yin, Bing, Hua, Yining, Zhou, Xuan, Rohanian, Omid, Thakur, Anshul, Clifton, Lei, Clifton, David A.

Reviewed 2026-10-27. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.