List your company

Mathematical Reasoning on MATH (Pass@1, Pass@32)

70.25Pass@1HRPO
23.491635.630847.7759.9092Jun 4, 2026

Evaluation results

MethodLinks
HRPOBackbone=Qwen2.5-7B-Instruct2026.0670.2588.2
TARPOBackbone=Qwen2.5-7B-Instruct2026.0670.2288.22
GRPOBackbone=Qwen2.5-7B-Instruct2026.0670.0588.02
CoTBackbone=Qwen2.5-7B-Instruct2026.0661.1687.3
HRPOBackbone=Qwen2.5-3B-Instruct2026.0659.9785.06
GRPOBackbone=Qwen2.5-3B-Instruct2026.0659.8684.72
TARPOBackbone=Qwen2.5-3B-Instruct2026.0659.7585.2
Pure LatentBackbone=Qwen2.5-7B-Instruct2026.0656.7868.66
CoTBackbone=Qwen2.5-3B-Instruct2026.0655.6485
Pure LatentBackbone=Qwen2.5-3B-Instruct2026.0655.4559.7
Entropy-RoutedBackbone=Qwen2.5-3B-Instruct2026.0654.7583.36
Entropy-RoutedBackbone=Qwen2.5-7B-Instruct2026.0654.0282.96
TARPOBackbone=Qwen2.5-1.5B-Instruct2026.0649.7182.28
HRPOBackbone=Qwen2.5-1.5B-Instruct2026.0648.9982.34
GRPOBackbone=Qwen2.5-1.5B-Instruct2026.0647.8281.22
CoTBackbone=Qwen2.5-1.5B-Instruct2026.0631.6678.9
Pure LatentBackbone=Qwen2.5-1.5B-Instruct2026.0625.2937.2
Entropy-RoutedBackbone=Qwen2.5-1.5B-Instruct2026.0625.2972.04