List your company

Claim-level Confidence Calibration on BeyondAIME

0.301SNR GainQwen3-4B-Instruct-confidence-min
0.006680.083090.15950.23591Dec 22, 2025

Evaluation results

MethodLinks
Qwen3-4B-Instruct-confidence-minAggregation=Minimum, Feature=Verbalized Confidence2025.120.30178.861.625.825.10.8153.7
Qwen3-4B-Instruct-confidence-prodAggregation=Product, Feature=Verbalized Confidence2025.120.18374.152.631.130.50.9349.7
o4-mini2025.120.06860.177.42121.10.88777
GLM-4.52025.120.02967.671.929.4261.02771.9
Qwen3-4B-Instruct2025.120.02761.756.345.641.91.80955.8
Gemini-2.5-Pro2025.120.01971.989.11310.40.44989.1
Qwen3-max2025.120.01871.788.913.210.60.46988.9