List your company

Question Answering on PubMedQA (Acc, TNFT, Tokens, Delay)

73.6AccuracyVanilla
69.64870.67471.772.726May 12, 2026

Evaluation results

MethodLinks
VanillaBackbone=Qwen3-4B2026.0573.6404.671,420.1421.73
OursBackbone=Qwen3-4B2026.0573.50648.9613.8
BaseBackbone=Qwen3-4B2026.0572.9389.61,323.7520.98
OursBackbone=Qwen3-1.7B2026.0570.50658.0110.96
VanillaBackbone=Qwen3-1.7B2026.0570.4404.671,422.9115.6
BaseBackbone=Qwen3-1.7B2026.0569.8389.61,263.9114.2