File size: 340 Bytes
ed1cff2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 | ---
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|
|