Mathematical Reasoning on AIME 25 (pass@1, pass@256)
13Pass@1F-CISPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| F-CISPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 13 | 64.6 | |
| DAPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 12 | 45.6 | |
| F-DAPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 11.5 | 52.9 | |
| F-CISPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 11.2 | 59.7 | |
| F-GRPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 10.5 | 57.9 | |
| CISPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 10.5 | 53.8 | |
| F-DAPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 10.3 | 61 | |
| F-GRPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 10.1 | 52.6 | |
| GRPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 9.9 | 58 | |
| DAPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 9.8 | 59.2 | |
| CISPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 9.7 | 59.8 | |
| GRPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 6.7 | 40.8 | |
| F-CISPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 2 | 24.5 | |
| F-DAPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 1.7 | 28.7 | |
| F-GRPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 1 | 29.5 | |
| DAPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 1 | 18.5 | |
| CISPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 1 | 25.4 | |
| GRPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 0.7 | 21.5 |