List your company

Medical Reasoning on NEJM (OOD)

68.3AccuracyPRA (Ours)
44.3850.5956.863.01Apr 10, 2026

Evaluation results

MethodLinks
PRA (Ours)Policy=Qwen3-4B-Instruct2026.0468.3
RAG + SCPolicy=Qwen3-4B-Instruct2026.0466.9
RAGPolicy=Qwen3-4B-Instruct2026.0465.7
CoT + SCPolicy=Qwen3-4B-Instruct2026.0463.2
CoTPolicy=Qwen3-4B-Instruct2026.0462.7
Direct + SCPolicy=Qwen3-4B-Instruct2026.0446.4
DirectPolicy=Qwen3-4B-Instruct2026.0445.3