List your company

Mathematical Reasoning on AIME 2025 (Mean@8)

59.2Mean@8DISA
1.095216.180131.26546.3499May 17, 2026

Evaluation results

MethodLinks
DISABackbone=Qwen3-4B-Base2026.0559.2
FlowRLBackbone=Qwen3-4B-Base2026.0553.3
GSPOBackbone=Qwen3-4B-Base2026.0552.1
GRPOBackbone=Qwen3-4B-Base2026.0540
DISABackbone=Qwen2.5-7B2026.0513.3
FlowRLBackbone=Qwen2.5-7B2026.059.17
GRPOBackbone=Qwen2.5-7B2026.057.92
VanillaBackbone=Qwen3-4B-Base2026.056.25
GSPOBackbone=Qwen2.5-7B2026.054.17
VanillaBackbone=Qwen2.5-7B2026.053.33