Boosting LLM Reasoning via Spontaneous Self-Correction

Zhao, Xutong, Xu, Tengyu, Wang, Xuewei · arXiv · 2025