List your company

Multiple-choice Question Reasoning on MMLU Pro (Acc, SL, SE)

78.57Accuracy (Acc)Cut2048
49.356456.940764.52572.1093Mar 23, 2026

Evaluation results

MethodLinks
Cut2048max tokens=20482026.0378.572,8392.77
ROM_CSCCounterfactual Self-Correction=true2026.0377.11,8184.24
EAT2026.0376.72,0523.74
Vanilla2026.0376.672,8422.7
ROM2026.0375.22,2993.27
Cut1024max tokens=10242026.0372.382,8392.55
L12026.0371.41,4015.1
Cut512max tokens=5122026.0350.482,8391.78