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

Utility Evaluation on MMLU (pass@1 Accuracy)

78Accuracy (pass@1)

SafeDecoding

28.0841.045466.96Aug 6, 2025
Updated 2mo ago

Evaluation Results

MethodLinks
2025.08
78
2025.08
74
2025.08
73
2025.08
73
2025.08
73
2025.08
73
2025.08
72
2025.08
69
2025.08
69
2025.08
69
2025.08
69
2025.08
67
2025.08
67
2025.08
66
2025.08
66
2025.08
65
2025.08
63
2025.08
59
2025.08
43
2025.08
42
2025.08
38
2025.08
30