Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Calibration on stream 2,000-question

0.08ECE

Verbalized + Temp Scaling

-1.54889.445620.4431.4344Mar 18, 2026
Updated 5d ago

Evaluation Results

MethodLinks
2026.03
0.080.2130.68164.40.089
2026.03
0.0830.2220.64364.60.08
2026.03
0.1080.2080.71664.60.103
2026.03
0.120.2110.71864.60.117
2026.03
0.2250.2580.68464.40.225
2026.03
524.158.7--
2026.03
5.625.454.8--
2026.03
6.522.369.4--
2026.03
8.322.264.3--
2026.03
10.120.776.2--
2026.03
1121.274.9--
2026.03
1125.152.1--
2026.03
11.72175.6--
2026.03
1221.171.8--
2026.03
14.125.965--
2026.03
15.422.767.5--
2026.03
1729.251--
2026.03
22.525.868.4--
2026.03
25.127.560--
2026.03
25.631.455.8--
2026.03
40.84156.6--