List your company

MCQA Suite (ARC-Challenge, ARC-Easy, HellaSwag, Winogrande, SciQ, PIQA)

43.6ARC-Challenge AccuracyComposite - Stretched
27.06431.35735.6539.943Oct 1, 2025

Evaluation results

MethodLinks
Composite - StretchedSize=8B2025.1043.671.870.466.591.378.570.4
Composite - StackedSize=8B2025.1042.771.570.664.89177.469.7
Llama 3.2Size=8B2025.1042.270.970.26690.576.569.3
Composite - StackedSize=1B2025.1033.559.556.759.483.273.160.9
Composite - StretchedSize=3B2025.1033.566.957.957.58472.362
Composite - StretchedSize=1B2025.1032.561.556.156.188.172.561.1
Composite - StackedSize=3B2025.1031.560.357.157.285.274.461
Composite - StretchedSize=350M2025.1030.555.45054.479.670.356.7
Llama 3.2Size=3B2025.1030.158.654.356.483.172.859.2
Llama 3.2Size=1B2025.103056.852.955.981.770.758
Composite - StackedSize=350M2025.1027.953.949.654.878.77156
Llama 3.2Size=350M2025.1027.753.249.254.178.370.155.4