List your company

Medical Question Answering on PubMedQA (Accuracy and Length Metrics)

0.807AccuracyAdaSR-HRPO
0.579240.638370.69750.75663Jun 12, 2026

Evaluation results

MethodLinks
AdaSR-HRPOModel=Qwen3-4B2026.060.807356.728114.512471.24
AdaSR-HRPOModel=Qwen3-1.7B2026.060.793414.206136.642550.848
AdaSR-GRPOModel=Qwen3-4B2026.060.792378.214111.842490.056
AdaSR-GRPOModel=Qwen3-1.7B2026.060.776433.573143.314576.887
StreamingThinker (SFT)Model=Qwen3-4B2026.060.756402.877116.106518.983
StreamingThinker (SFT)Model=Qwen3-1.7B2026.060.749414.737120.534535.271
Read-then-thinkModel=Qwen3-4B2026.060.615—331.835331.835
Read-then-thinkModel=Qwen3-1.7B2026.060.588—434.629434.629