Atlas

Benchmarks

← All benchmarks

MMLU Anatomy (Open Medical-LLM)

Science · 2024-04-19

The anatomy subset of MMLU, 135 questions, scored by the Open Medical-LLM Leaderboard under lm-evaluation-harness. Carried as its own family rather than pooled into a general MMLU score, because the leaderboard reports the medical subsets separately and pooling them would hide which clinical area a model is weak in.

Top models (higher is better)

ModelScore
GPT-480.0
Llama 3 8B68.9
GPT-3.5 Turbo 110665.9
Llama 3 8B Instruct62.2
Gemma 7B58.5
Solar 10.7B Instruct V1.057.8
Zephyr 7B Beta57.0
Mistral 7B v0.1 Base55.6
Qwen1.5 7B Chat54.8
Gemma 1.1 7B IT53.3
Qwen1.5 7B49.6
Gemma 7B IT48.1
Mistral 7B Instruct v0.145.2
Phi-1.539.3
Gemma 2B37.0
Falcon 7B25.2
Loading Atlas data…