Pre-trained LLMs Meet Sequential Recommenders: Efficient User-Centric Knowledge Distillation
For practitioners of sequential recommendation, this work offers a practical way to leverage LLM reasoning capabilities without incurring high inference costs, though the improvement is incremental over existing methods.
The paper proposes a knowledge distillation method that transfers user profiles generated by LLMs into sequential recommenders, improving recommendation quality without LLM inference at serving time. The method achieves competitive performance while maintaining inference efficiency comparable to traditional sequential models.
Sequential recommender systems have achieved significant success in modeling temporal user behavior but remain limited in capturing rich user semantics beyond interaction patterns. Large Language Models (LLMs) present opportunities to enhance user understanding with their reasoning capabilities, yet existing integration approaches create prohibitive inference costs in real time. To address these limitations, we present a novel knowledge distillation method that utilizes textual user profile generated by pre-trained LLMs into sequential recommenders without requiring LLM inference at serving time. The resulting approach maintains the inference efficiency of traditional sequential models while requiring neither architectural modifications nor LLM fine-tuning.