List your company

Multitask Language Understanding on MMLU Pro (Acc, Cost)

64.46AccuracyMACA
58.906460.348261.7963.2318May 25, 2026

Evaluation results

MethodLinks
MACABackbone=Qwen2.5-14B-Instruct2026.0564.462,412.5
DyLANBackbone=Qwen2.5-14B-Instruct2026.0564.0811,960.2
AgentPruneBackbone=Qwen2.5-14B-Instruct2026.0563.712,958.4
MaASBackbone=Qwen2.5-14B-Instruct2026.0563.223,147.9
CoTBackbone=Qwen2.5-14B-Instruct2026.0560.94286.7
VanillaBackbone=Qwen2.5-14B-Instruct2026.0559.12248.6