CoVE: Compressed Vocabulary Expansion Makes Better LLM-based Recommender Systems

2025-06-24

AI Summary

Existing LLM-based recommendation methods inadequately model user–item interaction sequences, resulting in limited sequential modeling capability. To address this, we propose the Compression-aware Vocabulary Expansion (CoVE) framework. CoVE explicitly captures sequential structure by assigning each item a unique identifier and expanding the LLM’s vocabulary accordingly. Concurrently, it introduces an embedding-layer compression mechanism that scales the vocabulary while constraining the growth of embedding parameters. This design preserves the LLM’s inherent sequential understanding capacity while substantially improving computational efficiency and scalability. Extensive experiments on multiple public recommendation benchmarks demonstrate that CoVE consistently outperforms state-of-the-art methods in accuracy, with significantly lower inference latency and reduced memory footprint—indicating strong potential for industrial deployment.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Computer Vision: Large Vision Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recommender systems play a pivotal role in providing relevant content to users. With the rapid development of large language models (LLMs), researchers have begun utilizing LLMs to build more powerful recommender systems. However, existing approaches that focus on aligning LLMs with recommendation tasks do not fully leverage their sequential information processing capabilities, leading to suboptimal performance. In this paper, we propose a novel system called compressed vocabulary expansion (CoVE). In CoVE, each item is assigned a unique ID within the expanded vocabulary. Our framework effectively capitalizes on sequence understanding abilities of LLMs, significantly enhancing their performance on recommendation tasks. Additionally, we compress the embedding layer, making CoVE practical for large-scale industrial applications. The effectiveness and performance of CoVE are demonstrated through comprehensive experiments on multiple recommendation datasets and comparisons with prior works. Our code can be found at https://github.com/HaochenZhang717/CoVE-official-Repo.

Problem

• Enhance LLM-based recommender systems' performance
• Leverage LLMs' sequential information processing capabilities
• Compress embedding layer for large-scale applications

Innovation

• Unique item IDs in expanded vocabulary
• Leverages LLMs' sequence understanding abilities
• Compressed embedding layer for scalability