Does increasing GPT-5.2 reasoning effort improve diagnosis accuracy enough to justify the token/latency cost? Ablation study on 897 paired medical cases.
benchmark openai diagnosis medical-ai ablation-study clinical-reasoning llm-evaluation mcnemar-test gpt-52 reasoning-effort medcasereasoning
-
Updated
Mar 25, 2026 - Python