On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models

On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models

来源：arXiv · arxiv.org