List your company

Mathematical Reasoning on AIME 25 (Avg@32, Avg)

52.5Avg@32DC-GRPO
-2.112.07526.2540.425May 27, 2026May 28, 2026May 30, 2026Jun 1, 2026Jun 3, 2026Jun 5, 2026Jun 7, 2026

Evaluation results

MethodLinks
DC-GRPOBackbone=Qwen3-8B, Precision=BF162026.0652.571
DC-GRPOBackbone=Qwen3-30B-A3B, Precision=BF162026.0649.171.8
GRPOBackbone=Qwen3-30B-A3B, Precision=BF162026.0643.669
GRPOBackbone=Qwen3-8B, Precision=BF162026.064067.3
OISDBackbone=Qwen3-4B2026.0536.3564.75
OISDBackbone=Qwen3-8B2026.0534.4866.72
BuPOBackbone=Qwen3-8B2026.0534.3866.36
GRPOBackbone=Qwen3-8B2026.0533.5464.23
BuPOBackbone=Qwen3-4B2026.0531.1558.51
GRPOBackbone=Qwen3-4B2026.0528.8555.08
PPOBackbone=Qwen3-4B2026.0527.655.22
RLOOBackbone=Qwen3-4B2026.0524.7954
VanillaBackbone=Qwen3-8B2026.0519.1748.49
Reinforce++Backbone=Qwen3-4B2026.0518.6545.03
VanillaBackbone=Qwen3-4B2026.0518.647.44
OISDBackbone=OctoThinker-8B-Long-Base2026.057.530.09
BuPOBackbone=OctoThinker-8B-Long-Base2026.056.7727.79
GRPOBackbone=OctoThinker-8B-Long-Base2026.052.1924.11
BuPOBackbone=OctoThinker-3B-Long-Base2026.050.4219.59
OISDBackbone=OctoThinker-3B-Long-Base2026.050.4219.91
GRPOBackbone=OctoThinker-3B-Long-Base2026.050.118.58
VanillaBackbone=OctoThinker-8B-Long-Base2026.050.13.75
VanillaBackbone=OctoThinker-3B-Long-Base2026.0501.68