Mathematical Reasoning on MATH 500 (pass@1, pass@16)
85.8Pass@1GRPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPOBackbone=Qwen3-8B-Base, Supervision Type=Labeled reward data2026.06 | 85.8 | 94.9 | |
| TTRL-CoCoVBackbone=Qwen3-8B-Base, Supervision Type=Label-free2026.06 | 82.5 | 94.6 | |
| GRPOBackbone=Qwen3-4B-Base, Supervision Type=Labeled reward data2026.06 | 81.6 | 91.9 | |
| TTRL-CoCoVBackbone=Qwen3-4B-Base, Supervision Type=Label-free2026.06 | 81.6 | 93.2 | |
| Co-rewardingBackbone=Qwen3-8B-Base, Supervision Type=Label-free2026.06 | 80.5 | 89.5 | |
| SCRLBackbone=Qwen3-8B-Base, Supervision Type=Label-free2026.06 | 77.8 | 87.9 | |
| Co-rewardingBackbone=Qwen3-4B-Base, Supervision Type=Label-free2026.06 | 77.5 | 86.2 | |
| TTRLBackbone=Qwen3-8B-Base, Supervision Type=Label-free2026.06 | 77 | 87.7 | |
| SCRLBackbone=Qwen3-4B-Base, Supervision Type=Label-free2026.06 | 75.3 | 88.1 | |
| TTRLBackbone=Qwen3-4B-Base, Supervision Type=Label-free2026.06 | 72.8 | 85.3 | |
| BaseBackbone=Qwen3-4B-Base, Supervision Type=Label-free2026.06 | 67.5 | 90.2 | |
| BaseBackbone=Qwen3-8B-Base, Supervision Type=Label-free2026.06 | 63.3 | 91.2 |