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

Preference Learning on Toy dataset 0% label noise (test)

99.6Accuracy

SimPO

57.37668.33879.390.262Oct 28, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.10
99.6
2025.10
99.6
2025.10
99.5
2025.10
99.3
2025.10
99.2
2025.10
99.1
2025.10
99.1
2025.10
98.6
2025.10
98.6
2025.10
98.6
2025.10
98.5
2025.10
98.3
2025.10
98.3
2025.10
98.1
2025.10
96.6
2025.10
96.6
2025.10
96.5
2025.10
96.5
2025.10
96.5
2025.10
96.5
2025.10
96
2025.10
96
2025.10
96
2025.10
95.5
2025.10
95.5
2025.10
93.4
2025.10
91
2025.10
91
2025.10
88.9
2025.10
88.9
2025.10
87.9
2025.10
87.9
2025.10
87.9
2025.10
86.1
2025.10
86.1
2025.10
85.7
2025.10
85.7
2025.10
84.6
2025.10
84.6
2025.10
84.6
2025.10
84.1
2025.10
84.1
2025.10
84.1
2025.10
84
2025.10
84
2025.10
83.2
2025.10
82.2
2025.10
82.2
2025.10
81.7
2025.10
81.7
2025.10
81.7
2025.10
81
2025.10
81
2025.10
80
2025.10
80
2025.10
77.7
2025.10
77.7
2025.10
77.7
2025.10
77.6
2025.10
77.6
2025.10
77.6
2025.10
76.2
2025.10
76.2
2025.10
76.2
2025.10
74.3
2025.10
74.3
2025.10
74.3
2025.10
71
2025.10
71
2025.10
71
2025.10
67.9
2025.10
67.9
2025.10
67.9
2025.10
59
2025.10
59
2025.10
59