Enhancing Large Language Model Based Sequential Recommender Systems with Pseudo Labels Reconstruction
Hyunsoo Na, Minseok Gang, Youngrok Ko, Jinseok Seol, Sang‐Goo Lee · 2024
Large language models (LLMs) are utilized in various studies, and have demonstrated potential to function independently as a recommendation model.However, training on useritem interaction sequences and additional textual information such as reviews often modifies the pre-trained weights of LLMs, diminishing their inherent strength in constructing and comprehending natural language sentences.In this study, we propose a reconstruction-based LLM recommendation model (ReLRec) that harnesses the feature extraction capability of LLMs, while preserving LLMs' sentence generation abilities.We reconstruct the user and item pseudo-labels generated from user reviews while training on sequential data, aiming to exploit the key features of both users and items.Experimental results demonstrate the efficacy of label reconstruction in sequential recommendation tasks.