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

Zero-shot Accuracy on MMLU, PIQA, ARC-E, ARC-C, Winogrande, and OBQA

63.89Avg. Accuracy (Zero-Shot)

Dense

32.627640.743848.8656.9762May 17, 2026
Updated 2mo ago

Evaluation Results

MethodLinks
2026.05
63.89
2026.05
57.95
2026.05
57.71
2026.05
57.5
2026.05
57.33
2026.05
56.82
2026.05
55.81
2026.05
55.09
2026.05
54.47
2026.05
54.44
2026.05
52.28
2026.05
50.81
2026.05
50.61
2026.05
50.39
2026.05
50.2
2026.05
49.95
2026.05
49.63
2026.05
49.1
2026.05
49.09
2026.05
48.87
2026.05
48.48
2026.05
45.19
2026.05
44.99
2026.05
44.93
2026.05
44.81
2026.05
44.56
2026.05
44.53
2026.05
43.88
2026.05
43.78
2026.05
43.41
2026.05
43.39
2026.05
43.17
2026.05
43.05
2026.05
42.59
2026.05
42.49
2026.05
42.42
2026.05
42.32
2026.05
42.29
2026.05
42.28
2026.05
42.1
2026.05
41.93
2026.05
41.71
2026.05
41.62
2026.05
41.53
2026.05
40.91
2026.05
40.41
2026.05
39.98
2026.05
38.94
2026.05
38.77
2026.05
38.73
2026.05
38.65
2026.05
38.33
2026.05
37.64
2026.05
37.29
2026.05
36.95
2026.05
36.78
2026.05
36.64
2026.05
34.66
2026.05
33.83