| license: mit | |
| # Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning | |
| The file `test.jsonl` is used to monitor the model performance as training proceeds. | |
| |Level |Source | | |
| |--------|--------| | |
| |1 |GSM8K | | |
| |2 |MATH-500| | |
| |3 |AMC-23 | | |
| |4 |AIME2024| | |
| |5 |AIME2025| | |