Mathematical Problem Solving on AIME 2025 (Pass@k)
64.7Pass@kGCPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GCPOk=64, Base Model=Qwen3-1.7B2026.05 | 64.7 | |
| GCPOk=32, Base Model=Qwen3-1.7B2026.05 | 59.8 | |
| GCPOk=16, Base Model=Qwen3-1.7B2026.05 | 52.5 | |
| DQOk=64, Base Model=Qwen3-1.7B2026.05 | 46.8 | |
| GCPOk=8, Base Model=Qwen3-1.7B2026.05 | 44.7 | |
| DQOk=32, Base Model=Qwen3-1.7B2026.05 | 42.9 | |
| GRPOk=64, Base Model=Qwen3-1.7B2026.05 | 42.4 | |
| GRPOk=32, Base Model=Qwen3-1.7B2026.05 | 38.5 | |
| GCPOk=4, Base Model=Qwen3-1.7B2026.05 | 38.2 | |
| DQOk=16, Base Model=Qwen3-1.7B2026.05 | 37.7 | |
| Basek=64, Base Model=Qwen3-1.7B2026.05 | 37.5 | |
| DQOk=8, Base Model=Qwen3-1.7B2026.05 | 34.2 | |
| GRPOk=16, Base Model=Qwen3-1.7B2026.05 | 34 | |
| Basek=32, Base Model=Qwen3-1.7B2026.05 | 31.4 | |
| DQOk=4, Base Model=Qwen3-1.7B2026.05 | 31.1 | |
| GCPOk=2, Base Model=Qwen3-1.7B2026.05 | 30.8 | |
| GRPOk=8, Base Model=Qwen3-1.7B2026.05 | 30.8 | |
| GRPOk=4, Base Model=Qwen3-1.7B2026.05 | 28.3 | |
| DQOk=2, Base Model=Qwen3-1.7B2026.05 | 26.3 | |
| Basek=16, Base Model=Qwen3-1.7B2026.05 | 24.8 | |
| GRPOk=2, Base Model=Qwen3-1.7B2026.05 | 23.9 | |
| Basek=8, Base Model=Qwen3-1.7B2026.05 | 20.5 | |
| Basek=4, Base Model=Qwen3-1.7B2026.05 | 15.7 | |
| Basek=2, Base Model=Qwen3-1.7B2026.05 | 11.4 |