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

Zero-shot Reasoning on ARC-e, PIQA, OpenbookQA, Winogrande, HellaSwag, and MathQA

57Average Accuracy

Baseline

0.8415.423044.58Apr 2, 2026Apr 14, 2026Apr 27, 2026May 10, 2026May 23, 2026Jun 5, 2026Jun 18, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
57
2026.06
57
2026.04
51
2026.04
50
2026.06
49.9
2026.04
49
2026.04
49
2026.06
48.8
2026.04
45
2026.04
44
2026.04
43
2026.06
41.6
2026.04
41
2026.06
40
2026.06
39.5
2026.04
38
2026.04
34
2026.04
34
2026.06
33.6
2026.06
33.3
2026.06
32.8
2026.06
32.4
2026.06
32.1
2026.06
31.8
2026.06
31.7
2026.06
31.5
2026.06
31.4
2026.06
31.3
2026.06
31.3
2026.06
31.3
2026.06
31.2
2026.06
31.2
2026.06
31.1
2026.04
30
2026.04
11
2026.04
10
2026.04
8
2026.04
6
2026.04
4
2026.04
3