List your company

Mathematical Reasoning on AIME 2025 (Accuracy and Token Costs)

40Sample CountSC
17.01622.98328.9534.917Feb 10, 2026

Evaluation results

MethodLinks
SCBackbone=Qwen2.5-3B2026.0240—35.66.67
SCBackbone=Qwen2.5-7B2026.0240—37.316.67
SCBackbone=Gemma3-4B2026.0240—164.116.67
ESCBackbone=Qwen2.5-3B2026.0239.83—37.86.67
DSCBackbone=Qwen2.5-3B2026.0237.94.935.16.67
ESCBackbone=Qwen2.5-7B2026.0237.64—37.416.67
ACBackbone=Qwen2.5-3B2026.0236.73—42.46.67
DSCBackbone=Qwen2.5-7B2026.0233.865.232.716.67
ESCBackbone=Gemma3-4B2026.0232.72—136.316.67
ACBackbone=Qwen2.5-7B2026.0232.58—39.316.67
ACTSCBackbone=Qwen2.5-3B2026.0230.13—29.410
DSCBackbone=Gemma3-4B2026.0228.9617.7119.816.67
ACBackbone=Gemma3-4B2026.0227.12—119.116.67
ACTSCBackbone=Qwen2.5-7B2026.0226.2—27.613.33
ACTSCBackbone=Gemma3-4B2026.0217.9—74.516.67