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

LLM evaluation human preference on PPE Human Preference track

0.282MSE / PPI

AIPW-EM

Updated 1mo ago

Evaluation Results

Method	Links
AIPW-EM 2026.04		0.282	2.8	94.5
LinearCal 2026.04		0.282	2.8	94.4
AutoCal 2026.04		0.282	2.7	94.4
MonoSpline 2026.04		0.282	2.7	94.2
IsoCal 2026.04		0.282	2.6	94
LinearCal 2026.04		0.282	2.8	94.4
AutoCal 2026.04		0.282	2.7	94.4
PPI++ 2026.04		0.283	2.5	94.6
Labeled-only 2026.04		0.289	0	94.5
Labeled-only 2026.04		0.289	0	94.5
AIPW-EM 2026.04		0.295	3.1	90.9
LinearCal 2026.04		0.295	3.1	90.9
LinearCal 2026.04		0.295	3.1	90.9
AutoCal 2026.04		0.297	2.8	90.8
MonoSpline 2026.04		0.297	2.5	90.3
IsoCal 2026.04		0.297	2.4	90.5
AutoCal 2026.04		0.297	2.8	90.8
PPI++ 2026.04		0.298	2.7	91.2
Labeled-only 2026.04		0.303	0	91.1
Labeled-only 2026.04		0.303	0	91.1
AIPW 2026.04		0.703	0.4	91.7
AIPW 2026.04		0.703	0.4	91.7
AIPW 2026.04		0.919	0.4	89.6
AIPW 2026.04		0.919	0.4	89.6
PPI 2026.04		1	0.4	89.7
PPI 2026.04		1	0.5	91.5
PPI 2026.04		1	0.4	89.7
PPI 2026.04		1	0.5	91.5