List your company

Language Understanding on OLMoE Downstream Evaluation Suite (MMLU, ARC-E, BoolQ, HellaSwag, PIQA, SciQ)

27.6MMLUL2R (SIPS)
22.50423.82725.1526.473Jan 29, 2026

Evaluation results

MethodLinks
L2R (SIPS)Training Tokens=10B, Scoring Geometry=SIPS2026.0127.643.861.740.662.371.543.4
L2R (Cosine)Training Tokens=10B, Scoring Geometry=Cosine2026.0124.942.261.142.26072.742
X-MoETraining Tokens=10B, Router=Cosine scoring2026.0124.646.96040.664.568.842.1
L2R (Dot)Training Tokens=10B, Scoring Geometry=Dot2026.0124.443.459.239.560.564.840.6
LinearTraining Tokens=10B, Router=Linear2026.0122.747.76038.761.166.840.5