List your company

Mathematical Reasoning on AIME 24/25/26 (val)

54.79Avg@16 AccuracyCPPO
7.05419.44731.8444.233Jun 9, 2026

Evaluation results

MethodLinks
CPPOBackbone=Qwen3-30B-A3B-Base2026.0654.79
DPPOBackbone=Qwen3-30B-A3B-Base2026.0649.23
TRM-AvgBackbone=Qwen3-30B-A3B-Base2026.0648.96
MinPROBackbone=Qwen3-30B-A3B-Base2026.0648.12
GRPOBackbone=Qwen3-30B-A3B-Base2026.0638.19
CPPOBackbone=Qwen3-1.7B2026.0631.88
CPPOBackbone=Qwen3-8B-Base2026.0631.11
MinPROBackbone=Qwen3-8B-Base2026.0629.72
CISPOBackbone=Qwen3-8B-Base2026.0629.58
DPPOBackbone=Qwen3-8B-Base2026.0628.89
CISPOBackbone=Qwen3-1.7B2026.0628.82
DPPOBackbone=Qwen3-1.7B2026.0628.19
TRM-AvgBackbone=Qwen3-8B-Base2026.0627.98
GRPOBackbone=Qwen3-1.7B2026.0627.91
MinPROBackbone=Qwen3-1.7B2026.0627.71
TRM-AvgBackbone=Qwen3-1.7B2026.0626.87
TRM-MaxBackbone=Qwen3-8B-Base2026.0626.73
TRM-MaxBackbone=Qwen3-1.7B2026.0625.21
GRPOBackbone=Qwen3-8B-Base2026.0623.96
TRM-MaxBackbone=Qwen3-30B-A3B-Base2026.0620.27
CPPOBackbone=Qwen3-1.7B-Base2026.0612.78
CISPOBackbone=Qwen3-1.7B-Base2026.0611.87
TRM-AvgBackbone=Qwen3-1.7B-Base2026.0611.7
MinPROBackbone=Qwen3-1.7B-Base2026.0611.04
DPPOBackbone=Qwen3-1.7B-Base2026.0610.9
TRM-MaxBackbone=Qwen3-1.7B-Base2026.069.72
GRPOBackbone=Qwen3-1.7B-Base2026.068.89