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

Reward Modeling on Skywork-Reward-Preference (test)

93.6Final Token Accuracy

TCRM

77.58481.74285.990.058Apr 24, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
93.688.70.020.85
2026.04
93.688.90.010.34
2026.04
93.560.10.992.1
2026.04
93.260.11.795.16
2026.04
92.651.19.1813.62
2026.04
92.586.30.041.05
2026.04
92.584.70.061.22
2026.04
92.371.70.082.88
2026.04
90.474.30.232.22
2026.04
90.350.48.1220.58
2026.04
90.3524.2710.68
2026.04
90620.343.88
2026.04
89.764.50.363.25
2026.04
89.778.10.261.68
2026.04
86.552.33.916.22
2026.04
86.554.90.643.65
2026.04
86.557.60.673.19
2026.04
84.479.80.0256.72
2026.04
84.263.50.642.18
2026.04
8452.60.623.24
2026.04
83.9485.3510.7
2026.04
83.479.704.88
2026.04
83.378014.67
2026.04
8377.808.73
2026.04
82.678.304.78
2026.04
79.875.306.98
2026.04
78.273.605.59