Chenheng Zhang

We rethink LLMs from the perspective of recommender systems and propose Language System, which exploits the model’s output distribution more efficiently and effectively.