Mathematical Reasoning on HMMT25 (Pass@1, Pass@128)
28Pass@1SRGen
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SRGenModel=Qwen3-32B2025.10 | 28 | — | |
| OXAFullModel Backbone=Qwen2.5-7B-Math, RLVR Stage=After RLVR (†)2026.03 | 26.6 | 66.7 | |
| OXAMLEModel Backbone=Qwen2.5-7B-Math, RLVR Stage=After RLVR (†)2026.03 | 26.5 | 63.3 | |
| Self-RefineModel=Qwen3-32B2025.10 | 25.3 | — | |
| OXAMLEModel Backbone=Qwen2.5-7B-Math, RLVR Stage=Before RLVR2026.03 | 24.5 | 80 | |
| SFTModel Backbone=Qwen2.5-7B-Math, RLVR Stage=Before RLVR2026.03 | 24.4 | 66.7 | |
| SFTModel Backbone=Qwen2.5-7B-Math, RLVR Stage=After RLVR (†)2026.03 | 23.6 | 70 | |
| CoTModel=Qwen3-32B2025.10 | 23.3 | — | |
| OXAFullModel Backbone=Qwen2.5-7B-Math, RLVR Stage=Before RLVR2026.03 | 23.1 | 70 | |
| SFTLPModel Backbone=Qwen2.5-7B-Math, RLVR Stage=After RLVR (†)2026.03 | 21.3 | 56.7 | |
| OXAMLEModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=After RLVR (†)2026.03 | 19.7 | 53.3 | |
| SFTLPModel Backbone=Qwen2.5-7B-Math, RLVR Stage=Before RLVR2026.03 | 19.3 | 63.3 | |
| SRGenModel=DS-R1-Qwen-7B2025.10 | 19.3 | — | |
| OXAFullModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=After RLVR (†)2026.03 | 19.1 | 66.7 | |
| OXAMLEModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=Before RLVR2026.03 | 18.5 | 60 | |
| OXAFullModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=Before RLVR2026.03 | 18.2 | 60 | |
| SRGenModel=DS-R1-Llama-8B2025.10 | 18 | — | |
| Self-RefineModel=DS-R1-Llama-8B2025.10 | 16 | — | |
| CoTModel=DS-R1-Qwen-7B2025.10 | 15.3 | — | |
| Self-RefineModel=DS-R1-Qwen-7B2025.10 | 14.7 | — | |
| CoTModel=DS-R1-Llama-8B2025.10 | 14 | — | |
| SFTModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=After RLVR (†)2026.03 | 13.8 | 56.7 | |
| SFTModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=Before RLVR2026.03 | 11.6 | 50 | |
| SFTLPModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=After RLVR (†)2026.03 | 9.5 | 46.7 | |
| SFTLPModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=Before RLVR2026.03 | 7.1 | 40 | |
| SRGenModel=Qwen2.5-Math-7B2025.10 | 3.3 | — | |
| CoTModel=Qwen2.5-Math-7B2025.10 | 1.3 | — | |
| BaseModel Backbone=Qwen2.5-7B-Math, RLVR Stage=Before RLVR2026.03 | 0.6 | 16.7 | |
| BaseModel Backbone=Qwen2.5-1.5B-Math, RLVR Stage=Before RLVR2026.03 | 0.5 | 26.7 | |
| Self-RefineModel=Qwen2.5-Math-7B2025.10 | 0 | — |