Mathematical Reasoning on AMC (pass@1, pass@256)
58.6Pass@1CISPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CISPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 58.6 | 95.7 | |
| CISPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 57.8 | 96.1 | |
| F-GRPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 56.2 | 96.3 | |
| F-DAPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 55.9 | 93.7 | |
| DAPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 54.5 | 95.2 | |
| F-DAPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 54.4 | 97 | |
| F-CISPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 53.8 | 99.1 | |
| DAPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 53.3 | 91.9 | |
| F-CISPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 53.3 | 97.1 | |
| GRPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 53.1 | 96.2 | |
| GRPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 52.9 | 87.3 | |
| F-GRPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 51.6 | 95.9 | |
| F-CISPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 34.1 | 82.6 | |
| DAPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 33.1 | 79.5 | |
| CISPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 32.9 | 79.1 | |
| F-DAPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 31.9 | 88.3 | |
| GRPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 30.5 | 88.2 | |
| F-GRPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 29.8 | 90.6 |