Math Reasoning on MATH (Pass@1)
86.58Pass@1SAS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SAS#Tok=17682026.04 | 86.58 | — | |
| DeepScaleR#Tok=28092026.04 | 86.36 | — | |
| ThinkPrune-4k#Tok=20902026.04 | 86.23 | — | |
| LAPO-I#Tok=22202026.04 | 86.04 | — | |
| GRPO-4K#Tok=19762026.04 | 85.09 | — | |
| L1-Max#Tok=14802026.04 | 83.6 | — | |
| GRPOextendedbackbone=Qwen2.5-7B2026.01 | 78 | — | |
| GRESObackbone=Qwen2.5-7B2026.01 | 77.8 | — | |
| LENSbackbone=Qwen2.5-7B2026.01 | 77.4 | — | |
| DAPObackbone=Qwen2.5-7B2026.01 | 77 | — | |
| GRPObackbone=Qwen2.5-7B2026.01 | 76.6 | — | |
| LENSbackbone=Qwen2.5-3B2026.01 | 68.2 | — | |
| DAPObackbone=Qwen2.5-3B2026.01 | 67.4 | — | |
| GRPOextendedbackbone=Qwen2.5-3B2026.01 | 66.2 | — | |
| GRESObackbone=Qwen2.5-3B2026.01 | 65.4 | — | |
| GRPObackbone=Qwen2.5-3B2026.01 | 65 | — | |
| Qwen2.5-7Bbackbone=Qwen2.5-7B2026.01 | 62.6 | — | |
| Qwen2.5-3Bbackbone=Qwen2.5-3B2026.01 | 56 | — |