Do LLMs Benefit from User and Item Embeddings in Recommendation Tasks?

2026-01-08

arXiv.org

AI Summary

This work addresses the limitation of existing large language models (LLMs) in recommendation tasks, which often fail to effectively integrate multi-item collaborative signals from user history, relying instead solely on textual semantics or isolated embeddings while neglecting structured interaction patterns. To bridge this gap, the authors propose a lightweight embedding projection module that maps user and item embeddings—learned via collaborative filtering—into the token space of the LLM, enabling joint modeling with textual inputs. This approach facilitates deep integration of structured collaborative signals into the language model without requiring substantial architectural modifications. Experimental results demonstrate that the method significantly outperforms pure text-based LLM baselines, achieving notable improvements across multiple recommendation metrics and underscoring the effectiveness and potential of combining structured interaction data with large language models for generative recommendation.

Technology Category

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

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: Large language models for search

Abstract

Large Language Models (LLMs) have emerged as promising recommendation systems, offering novel ways to model user preferences through generative approaches. However, many existing methods often rely solely on text semantics or incorporate collaborative signals in a limited manner, typically using only user or item embeddings. These methods struggle to handle multiple item embeddings representing user history, reverting to textual semantics and neglecting richer collaborative information. In this work, we propose a simple yet effective solution that projects user and item embeddings, learned from collaborative filtering, into the LLM token space via separate lightweight projector modules. A finetuned LLM then conditions on these projected embeddings alongside textual tokens to generate recommendations. Preliminary results show that this design effectively leverages structured user-item interaction data, improves recommendation performance over text-only LLM baselines, and offers a practical path for bridging traditional recommendation systems with modern LLMs.

Problem

• Large Language Models
• Recommendation Systems
• User Embeddings
• Item Embeddings
• Collaborative Filtering

Innovation

• Large Language Models
• Collaborative Filtering
• User-Item Embeddings
• Projection Module
• Recommendation Systems