medagentbench@1.0
Professional
Medicine
Assistants
MedAgentBench: 300 patient-specific clinically-derived tasks across 10 categories in a FHIR-compliant interactive healthcare environment.
Run this task
CLI:
inspect eval inspect_harbor/medagentbench_1_0 --model openai/gpt-5Python:
from inspect_ai import eval
from inspect_harbor import medagentbench_1_0
eval(medagentbench_1_0(), model="openai/gpt-5")Dataset information
| Harbor registry | medagentbench@1.0 |
| Inspect task | medagentbench_1_0 |
| Version | 1.0 |
| Samples | 300 |
| Paper | arxiv |
See Task Parameters for the parameter set shared across all Harbor tasks.