List your company

Mathematical Reasoning on AIME 2024 (Mean@32 accuracy)

75.2Mean@32 AccuracyHTPO
29.33641.24353.1565.057May 8, 2026

Evaluation results

MethodLinks
HTPOBackbone=Qwen3-8B-Instruct2026.0575.2
SAPOBackbone=Qwen3-8B-Instruct2026.0574
DAPOBackbone=Qwen3-8B-Instruct2026.0573.2
GSPOBackbone=Qwen3-8B-Instruct2026.0572.9
80/20-RuleBackbone=Qwen3-8B-Instruct2026.0572.9
BAPOBackbone=Qwen3-8B-Instruct2026.0572.8
GRPO†Backbone=Qwen3-8B-Instruct2026.0571.3
HTPOBackbone=Qwen3-8B-Base2026.0541.5
SAPOBackbone=Qwen3-8B-Base2026.0540.3
GSPOBackbone=Qwen3-8B-Base2026.0539.2
BAPOBackbone=Qwen3-8B-Base2026.0535.6
80/20-RuleBackbone=Qwen3-8B-Base2026.0534.3
DAPOBackbone=Qwen3-8B-Base2026.0532.9
GRPO†Backbone=Qwen3-8B-Base2026.0531.1