HealthBench-Hard

HealthBench-Hard is a benchmark used to evaluate the alignment of large language models with clinician preferences in healthcare contexts.

Papers using HealthBench-Hard (3)

ClinAlign: Scaling Healthcare Alignment from Clinician Preference2026InfiMed-ORBIT: Aligning LLMs on Open-Ended Complex Tasks via Rubric-Based Incremental Training2025InfiMed-ORBIT: Aligning LLMs on Open-Ended Complex Tasks via Rubric-Based Incremental Training2025