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

Zero-shot Reasoning on (ARC-C, BoolQ, HellaSwag, MathQA, MMLU, Race)

50.43ARC-C Accuracy

Base

19.635627.630335.62543.6197Oct 7, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.10
50.4383.6181.0835.7159.6440.8658.56
2025.10
43.3477.717628.0741.8239.5251.08
2025.10
37.7166.2162.2629.2528.536.9443.48
2025.10
36.5267.4360.9625.7627.4835.0242.2
2025.10
36.3565.8159.7428.3827.9935.542.3
2025.10
35.1567.7459.628.5428.5335.542.51
2025.10
34.3963.0959.1925.3326.234.5540.46
2025.10
33.768.0159.5524.9626.6433.9741.14
2025.10
23.6356.1243.4222.6523.2928.6132.62
2025.10
22.9543.2141.7322.2423.3428.830.83
2025.10
22.8742.5741.7422.3823.5427.8530.16
2025.10
21.6738.6237.8722.1823.3827.0828.47
2025.10
21.4239.0239.1222.3523.4427.4628.8
2025.10
20.8238.4137.8722.7523.2827.1828.39