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

Zero-shot Reasoning on ARC-C, ARC-E, HellaSwag, LAMBADA, OpenBookQA, PIQA, and WinoGrande

36.9ARC-C Accuracy

H-Net 1.3B (ours)

27.74830.12432.534.876May 28, 2026
Updated 5d ago

Evaluation Results

MethodLinks
2026.05
36.960.554.944.437.471.556.251.7
2026.05
34.458.75543.438.470.756.851.1
2026.05
33.961.252.442.137.271.756.150.7
2026.05
33.460.152.340.939.270.955.750.4
2026.05
33.160.34937.334.870.254.748.5
2026.05
28.153.738.326.732.464.55142.1