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

Zero-shot Evaluation on ARC (c/e), OBQA, WinoGrande, PIQA, HellaSwag, MathQA, RTE, and BoolQ

71.08Mean Accuracy

Original (Dense)

44.24851.21458.1865.146Mar 9, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.03
71.08
2026.03
71.08
2026.03
71.08
2026.03
70.04
2026.03
69.94
2026.03
69.94
2026.03
68.79
2026.03
68.79
2026.03
67.65
2026.03
67.46
2026.03
66.89
2026.03
66.63
2026.03
66.63
2026.03
66.35
2026.03
65.96
2026.03
65.96
2026.03
65.77
2026.03
64.82
2026.03
64.71
2026.03
64.21
2026.03
64.13
2026.03
63.35
2026.03
62.77
2026.03
62.41
2026.03
62.28
2026.03
62.28
2026.03
62.28
2026.03
62.14
2026.03
61.84
2026.03
61.7
2026.03
61.7
2026.03
61.2
2026.03
61.03
2026.03
60.62
2026.03
60.51
2026.03
60.38
2026.03
60.05
2026.03
59.03
2026.03
58.97
2026.03
58.58
2026.03
58.2
2026.03
58.18
2026.03
57.7
2026.03
57.13
2026.03
56.7
2026.03
56.07
2026.03
54.94
2026.03
54.75
2026.03
54.32
2026.03
53.16
2026.03
52.31
2026.03
52.17
2026.03
52.02
2026.03
52.02
2026.03
51.62
2026.03
51.38
2026.03
49.71
2026.03
46.72
2026.03
45.28