List your company
ResearchFormerly WizWand

Mathematical Reasoning on AIME (Solved Rate %)

83.33Solved Rate (%)Self-Refine
14.9532.702550.45568.2075May 21, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Self-RefineModel=Qwen3 32B think, Context Window=32k2025.0583.33
ICRLModel=Qwen3 32B think, Context Window=32k2025.0580
ReflexionModel=Qwen3 32B think, Context Window=32k2025.0570
BaseModel=Qwen3 32B think, Context Window=32k2025.0566.58
ICRLModel=Qwen3 32B, Context Window=32k2025.0546.66
Self-RefineModel=Qwen3 32B, Context Window=32k2025.0543.33
ReflexionModel=Phi-4, Context Window=16k2025.0540
ICRLModel=Phi-4, Context Window=16k2025.0540
ICRLModel=Llama 4 Maverick, Context Window=32k2025.0535
ReflexionModel=Qwen3 32B, Context Window=32k2025.0533.33
Self-RefineModel=Phi-4, Context Window=16k2025.0533.33
ReflexionModel=Llama 4 Maverick, Context Window=32k2025.0523.33
BaseModel=Qwen3 32B, Context Window=32k2025.0522.54
Self-RefineModel=Llama 4 Maverick, Context Window=32k2025.0520
BaseModel=Phi-4, Context Window=16k2025.0520
BaseModel=Llama 4 Maverick, Context Window=32k2025.0517.58