List your company

Question Answering on PubMedQA (Accuracy and ECE)

82.2AccuracyGemini 3 Pro
50.79258.94667.175.254Dec 25, 2025Jan 11, 2026Jan 28, 2026Feb 14, 2026Mar 3, 2026Mar 20, 2026Apr 6, 2026

Evaluation results

MethodLinks
Gemini 3 ProTemperature=default2026.0482.2—
Gemini 3 FlashTemperature=default2026.0480.8—
HPS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=High-PPL Sampling2025.1278.6—
DOS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=Distance-to-Optimum Selection2025.1277.4—
MedGemma 1Model Scale=27B, Temperature=0.02026.0477.2—
RS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=Random Data Sampling2025.1276.7—
Qwen3-14B-BaseBackbone=Qwen3-14B-Base, CPT Strategy=None2025.1276.6—
LPS-CPTBackbone=Qwen3-14B-Base, CPT Strategy=Low-PPL Sampling2025.1276.4—
Qwen3 VLModel Scale=4B, Temperature=default2026.0474.8—
MedGemma 1Model Scale=4B, Temperature=0.02026.0473.4—
MedGemma 1.5Model Scale=4B, Temperature=0.02026.0467.6—
UAT-LITESeed=Single2026.02640.103
BaselineSeed=Single2026.02620.114
Temperature ScalingSeed=Single2026.02620.128
Deep EnsembleEnsemble Size=5, Seed=Single2026.02520.051