Mathematical Reasoning on AIME 24 (pass@1 accuracy, average pass@1 accuracy)
21.9Pass@1 AccuracySCRL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SCRLBackbone=Llama-3.1-8B-Instruct2026.03 | 21.9 | 29 | |
| ETMRBackbone=Llama-3.1-8B-Instruct2026.03 | 16.9 | 26.2 | |
| RESTRAINBackbone=Llama-3.1-8B-Instruct2026.03 | 16.7 | 28.4 | |
| TTRLBackbone=Llama-3.1-8B-Instruct2026.03 | 10 | 21.2 | |
| TTRLBackbone=Llama-3.2-1B-Instruct2026.03 | 6.7 | 17.3 | |
| SCRLBackbone=Llama-3.2-1B-Instruct2026.03 | 6.7 | 23.2 | |
| Llama-3.1-8B-InstructBackbone=Llama-3.1-8B-Instruct2026.03 | 4.6 | 14 | |
| COMPASSBackbone=Llama-3.2-1B-Instruct2026.03 | 3.5 | 16.1 | |
| Llama-3.2-1B-InstructBackbone=Llama-3.2-1B-Instruct2026.03 | 1.5 | 13.1 |