List your company

Mathematical Reasoning on HMMT25 (avg@12 accuracy)

48.1Avg@12 AccuracyOPSD
19.18826.69434.241.706May 27, 2026

Evaluation results

MethodLinks
OPSDModel=Qwen3-8B2026.0548.1
SGSDModel=Qwen3-8B2026.0546.9
SGSDModel=Qwen3-4B2026.0546.7
OPSDModel=Qwen3-4B2026.0546.1
GRPOModel=Qwen3-8B2026.0546.1
GRPOModel=Qwen3-4B2026.0545.3
BaseModel=Qwen3-8B2026.0545
GRPO+SkillModel=Qwen3-8B2026.0545
GRPO+SkillModel=Qwen3-4B2026.0544.7
OPSD+SkillModel=Qwen3-4B2026.0544.7
OPSD+SkillModel=Qwen3-8B2026.0543.9
BaseModel=Qwen3-4B2026.0543.3
Base+SkillModel=Qwen3-4B2026.0542.2
Base+SkillModel=Qwen3-8B2026.0540.8
SGSDModel=Qwen3-1.7B2026.0529.7
OPSD+SkillModel=Qwen3-1.7B2026.0527.2
OPSDModel=Qwen3-1.7B2026.0526.9
GRPOModel=Qwen3-1.7B2026.0525.8
GRPO+SkillModel=Qwen3-1.7B2026.0525.8
BaseModel=Qwen3-1.7B2026.0524.2
Base+SkillModel=Qwen3-1.7B2026.0520.3