Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.02240 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914527585501184 |
|---|---|
| author | Liu, Ruoqi Mohiuddin, Imran Q. Schoeffler, Austin J. Renduchintala, Kavita Nayak, Ashwin Vemu, Prasantha L. Vedak, Shivam C. Black, Kameron C. Havlik, John L. Ogunmola, Isaac Ma, Stephen P. Dhatt, Roopa Chen, Jonathan H. |
| author_facet | Liu, Ruoqi Mohiuddin, Imran Q. Schoeffler, Austin J. Renduchintala, Kavita Nayak, Ashwin Vemu, Prasantha L. Vedak, Shivam C. Black, Kameron C. Havlik, John L. Ogunmola, Isaac Ma, Stephen P. Dhatt, Roopa Chen, Jonathan H. |
| contents | We introduce PhysicianBench, a benchmark for evaluating LLM agents on physician tasks grounded in real clinical setting within electronic health record (EHR) environments. Existing medical agent benchmarks primarily focus on static knowledge recall, single-step atomic actions, or action intent without verifiable execution against the environment. As a result, they fail to capture the long-horizon, composite workflows that characterize real clinical systems. PhysicianBench comprises 100 long-horizon tasks adapted from real consultation cases between primary care and subspecialty physicians, with each task independently reviewed by a separate panel of physicians. Tasks are instantiated in an EHR environment with real patient records and accessed through the same standard APIs used by commercial EHR vendors. Tasks span 21 specialties (e.g., cardiology, endocrinology, oncology, psychiatry) and diverse workflow types (e.g., diagnosis interpretation, medication prescribing, treatment planning), requiring an average of 27 tool calls per task. Solving each task requires retrieving data across encounters, reasoning over heterogeneous clinical information, executing consequential clinical actions, and producing clinical documentation. Each task is decomposed into structured checkpoints (670 in total across the benchmark) capturing distinct stages of completion graded by task-specific scripts with execution-grounded verification. Across 13 proprietary and open-source LLM agents, the best-performing model achieves only 46% success rate (pass@1), while open-source models reach at most 19%, revealing a substantial gap between current agent capabilities and the demands of real-world clinical workflows. PhysicianBench provides a realistic and execution-grounded benchmark for measuring progress toward autonomous clinical agents. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_02240 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | PhysicianBench: Evaluating LLM Agents in Real-World EHR Environments Liu, Ruoqi Mohiuddin, Imran Q. Schoeffler, Austin J. Renduchintala, Kavita Nayak, Ashwin Vemu, Prasantha L. Vedak, Shivam C. Black, Kameron C. Havlik, John L. Ogunmola, Isaac Ma, Stephen P. Dhatt, Roopa Chen, Jonathan H. Artificial Intelligence We introduce PhysicianBench, a benchmark for evaluating LLM agents on physician tasks grounded in real clinical setting within electronic health record (EHR) environments. Existing medical agent benchmarks primarily focus on static knowledge recall, single-step atomic actions, or action intent without verifiable execution against the environment. As a result, they fail to capture the long-horizon, composite workflows that characterize real clinical systems. PhysicianBench comprises 100 long-horizon tasks adapted from real consultation cases between primary care and subspecialty physicians, with each task independently reviewed by a separate panel of physicians. Tasks are instantiated in an EHR environment with real patient records and accessed through the same standard APIs used by commercial EHR vendors. Tasks span 21 specialties (e.g., cardiology, endocrinology, oncology, psychiatry) and diverse workflow types (e.g., diagnosis interpretation, medication prescribing, treatment planning), requiring an average of 27 tool calls per task. Solving each task requires retrieving data across encounters, reasoning over heterogeneous clinical information, executing consequential clinical actions, and producing clinical documentation. Each task is decomposed into structured checkpoints (670 in total across the benchmark) capturing distinct stages of completion graded by task-specific scripts with execution-grounded verification. Across 13 proprietary and open-source LLM agents, the best-performing model achieves only 46% success rate (pass@1), while open-source models reach at most 19%, revealing a substantial gap between current agent capabilities and the demands of real-world clinical workflows. PhysicianBench provides a realistic and execution-grounded benchmark for measuring progress toward autonomous clinical agents. |
| title | PhysicianBench: Evaluating LLM Agents in Real-World EHR Environments |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2605.02240 |