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

Zero-shot Commonsense Reasoning on ARC-Easy, ARC-Challenge, and HellaSwag Average

68.6Zero-Shot Accuracy

BF16

30.01640.03350.0560.067Jun 9, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
68.6
2026.06
62
2026.06
57.7
2026.06
56.5
2026.06
54.9
2026.06
52.6
2026.06
44
2026.06
41.4
2026.06
40.2
2026.06
38.1
2026.06
37.5
2026.06
37.1
2026.06
36.8
2026.06
36.3
2026.06
36.1
2026.06
35.5
2026.06
34.9
2026.06
34.8
2026.06
34.7
2026.06
34.4
2026.06
34.4
2026.06
33.9
2026.06
33.8
2026.06
33.7
2026.06
33.5
2026.06
33.4
2026.06
33.3
2026.06
33.3
2026.06
33.2
2026.06
33.1
2026.06
33
2026.06
33
2026.06
33
2026.06
32.9
2026.06
32.9
2026.06
32.9
2026.06
32.8
2026.06
32.7
2026.06
32.7
2026.06
32.7
2026.06
32.5
2026.06
32.5
2026.06
32.5
2026.06
32.4
2026.06
32.3
2026.06
32.3
2026.06
32.3
2026.06
32.2
2026.06
32.1
2026.06
32.1
2026.06
32.1
2026.06
32.1
2026.06
32
2026.06
31.9
2026.06
31.8
2026.06
31.8
2026.06
31.6
2026.06
31.6
2026.06
31.6
2026.06
31.5