List your company

Mathematical Reasoning on MATH 500 (Accuracy, Latency)

96.77AccuracyQwen3-8B
69.345276.465183.58590.7049May 13, 2026

Evaluation results

MethodLinks
Qwen3-8BBackbone=Qwen3-8B2026.0596.77352.3
Full BaselineBackbone=Qwen3-14B2026.0596.52355.5
FlowCompileSelection Priority=Accuracy-priority2026.0595.27185.5
Qwen3-4BBackbone=Qwen3-4B2026.0593.78187.4
FlowCompileSelection Priority=Latency-priority2026.0585.3231.9
KNN Router2026.0578.6182
MaASBackbone=Qwen3-4B2026.0571.8255
MaASBackbone=Qwen3-8B2026.0570.487.1