Implicit Reasoning for Large Language Model-based Generative Recommendation

The article introduces PauseRec, a novel implicit reasoning paradigm designed for Generative Recommendation (GR) using Large Language Models (LLMs). It addresses limitations of existing explicit reasoning methods, achieving up to 6.22% better performance, a 65% reduction in training costs, and a 71.3% increase in inference speed. This advancement is significant for practitioners as it provides a more efficient approach to leveraging LLMs for GR, minimizing the need for complex multi-stage reasoning pipelines.