List your company

Massive Multitask Language Understanding on MMLU Pro (Accuracy, Gain, Wins vs AIM)

67.84AccuracyGCWM
25.236.2747.3458.41May 10, 2026

Evaluation results

MethodLinks
GCWMScale=14B2026.0567.841.239
OPCMScale=14B2026.0566.61——
GCWMScale=8B2026.0563.660.748
AIMMergingScale=8B2026.0562.92——
GCWMScale=4B2026.0559.431.1914
OPCMScale=4B2026.0558.24——
GCWMScale=1.7B2026.0543.451.6112
AIMMergingScale=1.7B2026.0541.84——
GCWMScale=0.6B2026.0527.140.37
OPCMScale=0.6B2026.0526.84——