dorsal/arxiv
View SchemaMLB: A Scenario-Driven Benchmark for Evaluating Large Language Models in Clinical Applications
| Authors | Qing He, Dongsheng Bi, Jianrong Lu, Minghui Yang, Zixiao Chen, Jiacheng Lu, Jing Chen, Nannan Du, Xiao Cu, Sijing Wu, Peng Xiang, Yinyin Hu, Yi Guo, Chunpu Li, Shaoyang Li, Zhuo Dong, Ming Jiang, Shuai Guo, Liyun Feng, Jin Peng, Jian Wang, Jinjie Gu, Junwei Liu |
|---|---|
| Categories | |
| ArXiv ID | 2601.06193vv1 |
| URL | https://arxiv.org/abs/2601.06193 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
The proliferation of Large Language Models (LLMs) presents transformative potential for healthcare, yet practical deployment is hindered by the absence of frameworks that assess real-world clinical utility. Existing benchmarks test static knowledge, failing to capture the dynamic, application-oriented capabilities required in clinical practice. To bridge this gap, we introduce a Medical LLM Benchmark MLB, a comprehensive benchmark evaluating LLMs on both foundational knowledge and scenario-based reasoning. MLB is structured around five core dimensions: Medical Knowledge (MedKQA), Safety and Ethics (MedSE), Medical Record Understanding (MedRU), Smart Services (SmartServ), and Smart Healthcare (SmartCare). The benchmark integrates 22 datasets (17 newly curated) from diverse Chinese clinical sources, covering 64 clinical specialties. Its design features a rigorous curation pipeline involving 300 licensed physicians. Besides, we provide a scalable evaluation methodology, centered on a specialized judge model trained via Supervised Fine-Tuning (SFT) on expert annotations. Our comprehensive evaluation of 10 leading models reveals a critical translational gap: while the top-ranked model, Kimi-K2-Instruct (77.3% accuracy overall), excels in structured tasks like information extraction (87.8% accuracy in MedRU), performance plummets in patient-facing scenarios (61.3% in SmartServ). Moreover, the exceptional safety score (90.6% in MedSE) of the much smaller Baichuan-M2-32B highlights that targeted training is equally critical. Our specialized judge model, trained via SFT on a 19k expert-annotated medical dataset, achieves 92.1% accuracy, an F1-score of 94.37%, and a Cohen's Kappa of 81.3% for human-AI consistency, validating a reproducible and expert-aligned evaluation protocol. MLB thus provides a rigorous framework to guide the development of clinically viable LLMs.
{
"annotation_id": "11e45ccf-b086-4246-9a85-30263efbb354",
"date_created": "2026-02-17T05:53:08.670000Z",
"date_modified": "2026-02-17T05:53:08.670000Z",
"file_hash": "1debc8e3632e987d5519ed9f37d70c5354d9790c4e97c48357d488f49fb9a347",
"private": false,
"record": {
"abstract": "The proliferation of Large Language Models (LLMs) presents transformative potential for healthcare, yet practical deployment is hindered by the absence of frameworks that assess real-world clinical utility. Existing benchmarks test static knowledge, failing to capture the dynamic, application-oriented capabilities required in clinical practice. To bridge this gap, we introduce a Medical LLM Benchmark MLB, a comprehensive benchmark evaluating LLMs on both foundational knowledge and scenario-based reasoning. MLB is structured around five core dimensions: Medical Knowledge (MedKQA), Safety and Ethics (MedSE), Medical Record Understanding (MedRU), Smart Services (SmartServ), and Smart Healthcare (SmartCare). The benchmark integrates 22 datasets (17 newly curated) from diverse Chinese clinical sources, covering 64 clinical specialties. Its design features a rigorous curation pipeline involving 300 licensed physicians. Besides, we provide a scalable evaluation methodology, centered on a specialized judge model trained via Supervised Fine-Tuning (SFT) on expert annotations. Our comprehensive evaluation of 10 leading models reveals a critical translational gap: while the top-ranked model, Kimi-K2-Instruct (77.3% accuracy overall), excels in structured tasks like information extraction (87.8% accuracy in MedRU), performance plummets in patient-facing scenarios (61.3% in SmartServ). Moreover, the exceptional safety score (90.6% in MedSE) of the much smaller Baichuan-M2-32B highlights that targeted training is equally critical. Our specialized judge model, trained via SFT on a 19k expert-annotated medical dataset, achieves 92.1% accuracy, an F1-score of 94.37%, and a Cohen\u0027s Kappa of 81.3% for human-AI consistency, validating a reproducible and expert-aligned evaluation protocol. MLB thus provides a rigorous framework to guide the development of clinically viable LLMs.",
"arxiv_id": "2601.06193",
"authors": [
"Qing He",
"Dongsheng Bi",
"Jianrong Lu",
"Minghui Yang",
"Zixiao Chen",
"Jiacheng Lu",
"Jing Chen",
"Nannan Du",
"Xiao Cu",
"Sijing Wu",
"Peng Xiang",
"Yinyin Hu",
"Yi Guo",
"Chunpu Li",
"Shaoyang Li",
"Zhuo Dong",
"Ming Jiang",
"Shuai Guo",
"Liyun Feng",
"Jin Peng",
"Jian Wang",
"Jinjie Gu",
"Junwei Liu"
],
"categories": [
"cs.LG",
"cs.AI",
"cs.CL"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "MLB: A Scenario-Driven Benchmark for Evaluating Large Language Models in Clinical Applications",
"url": "https://arxiv.org/abs/2601.06193",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "0ccdf628-bbe3-49c0-8289-e29a5b0789f6",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}