Medmarks introduces 30 open benchmarks for medical LLM tasks and evaluates 61 models, finding frontier reasoning models lead while medically fine-tuned ones outperform generalists and all show answer-order bias.
TaskMCQA in English and Spanish
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks
Medmarks introduces 30 open benchmarks for medical LLM tasks and evaluates 61 models, finding frontier reasoning models lead while medically fine-tuned ones outperform generalists and all show answer-order bias.