Pre-trained LLMs Meet Sequential Recommenders: Efficient User-Centric Knowledge Distillation

2026-04-23

AI Summary

This work addresses the challenge that existing sequential recommendation systems struggle to capture deep user semantics, while directly integrating large language models (LLMs) incurs prohibitive online inference costs. To overcome this, the authors propose an efficient knowledge distillation approach that leverages a pretrained LLM offline to generate textual user profiles and injects their semantic knowledge into a lightweight sequential recommender via user-centric knowledge distillation. Notably, this method requires neither online LLM invocation, architectural modifications to the base recommender, nor fine-tuning of the LLM. By preserving the original inference efficiency, it substantially enhances the model’s semantic understanding of user behavior and overall recommendation performance, thereby achieving, for the first time, an efficient, fine-tuning-free, and architecture-agnostic semantic-enhanced sequential recommendation framework.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

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.

Problem

• sequential recommender systems
• Large Language Models
• user semantics
• inference cost
• knowledge distillation

Innovation

• knowledge distillation
• large language models
• sequential recommendation
• user profiling
• inference efficiency