Triplets Better Than Pairs: Towards Stable and Effective Self-Play Fine-Tuning for LLMs

Open in new window