Math Reasoning on AIME 2024 (Pass@1, AES)
39.79Pass@1SAS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SAS#Tok=48762026.04 | 39.79 | — | |
| GRPO-4K#Tok=52822026.04 | 38.75 | — | |
| ThinkPrune-4k#Tok=54682026.04 | 36.04 | — | |
| LAPO-I#Tok=56272026.04 | 34.58 | — | |
| DeepScaleR#Tok=67552026.04 | 33.75 | — | |
| L1-Max#Tok=21582026.04 | 25.63 | — |