List your company

Mathematical Reasoning on AIME 26 (val)

75.2Accuracy (%)Base
54.60859.95465.370.646Apr 10, 2026

Evaluation results

MethodLinks
BaseBackbone=Qwen3-32B, Generations=642026.0475.23.2426.7
ControlBackbone=Qwen3-32B, Generations=642026.0474.13.8335.2
MEMENTOBackbone=Qwen3-32B, Generations=642026.0472.61.6714
BaseBackbone=Phi-4-r (14B), Generations=642026.0471.72.6528.8
ControlBackbone=Phi-4-r (14B), Generations=642026.0469.83.0428.6
BaseBackbone=Olmo 3 (7B), Generations=642026.0467.93.9550.8
MEMENTOBackbone=Phi-4-r (14B), Generations=642026.0467.61.1711.3
BaseBackbone=Qwen3-8B, Generations=642026.0466.82.4125.3
Mem. + RLBackbone=Qwen3-8B, Generations=642026.0464.91.4514.9
ControlBackbone=Qwen3-8B, Generations=642026.0464.72.5928.3
ControlBackbone=Olmo 3 (7B), Generations=642026.0459.83.5137.1
MEMENTOBackbone=Qwen3-8B, Generations=642026.0457.31.029.7
MEMENTOBackbone=Olmo 3 (7B), Generations=642026.0455.43.2137.8