1 paper
Quanquan Li, Hongbo Zhang, Yihe Chi +9
Token-level collaboration allows a large language model (LLM) to assist a small language model (SLM) when their predictions diverge. Existing methods either use LLM-generated inter…