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

General Language Understanding on LAMBADA, ARC-C, HellaSwag, BoolQ, MMLU, and GSM8K

74.35Accuracy

BF16-Baseline

17.399632.184846.9761.7552Jun 22, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
74.35
2026.06
74.34
2026.06
74.19
2026.06
74.02
2026.06
73.91
2026.06
73.91
2026.06
73.81
2026.06
73.74
2026.06
73.7
2026.06
73.69
2026.06
73.36
2026.06
73.36
2026.06
73.33
2026.06
73.27
2026.06
73.24
2026.06
73.19
2026.06
73.14
2026.06
72.99
2026.06
72.86
2026.06
72.85
2026.06
72.84
2026.06
72.82
2026.06
72.75
2026.06
72.5
2026.06
72.34
2026.06
72.25
2026.06
72.18
2026.06
71.98
2026.06
71.79
2026.06
71.77
2026.06
71.75
2026.06
71.6
2026.06
71.44
2026.06
71.34
2026.06
70.94
2026.06
69.89
2026.06
69.28
2026.06
69.06
2026.06
67.64
2026.06
66.35
2026.06
65.45
2026.06
65.03
2026.06
64.67
2026.06
64.59
2026.06
63.4
2026.06
63.27
2026.06
63
2026.06
61.37
2026.06
59.19
2026.06
56.05
2026.06
50.64
2026.06
44.79
2026.06
33.8
2026.06
31.44
2026.06
22.22
2026.06
19.59