Large Language Models Are Universal Recommendation Learners

2025-02-05

AI Summary

To address the poor generalizability and high architectural design cost of task-specific models in real-world recommender systems, this paper investigates the feasibility of large language models (LLMs) as universal recommendation learners. We propose three key techniques: (1) a multimodal fusion module to enrich item semantic representations; (2) a sequence-to-set modeling paradigm that improves candidate generation efficiency; and (3) industrial-oriented prompt engineering strategies that significantly enhance LLMs’ instruction following and ranking capabilities for recommendation tasks. Extensive experiments on large-scale industrial datasets demonstrate that our approach achieves performance on par with dedicated expert models across multiple recommendation tasks—including click-through rate prediction, multi-objective ranking, and cold-start recommendation. This work provides the first systematic validation of LLMs as unified, lightweight, and scalable recommendation learners, establishing their practical viability and effectiveness in production settings.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Data Mining & Knowledge Management: Recommender Systems

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
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In real-world recommender systems, different tasks are typically addressed using supervised learning on task-specific datasets with carefully designed model architectures. We demonstrate that large language models (LLMs) can function as universal recommendation learners, capable of handling multiple tasks within a unified input-output framework, eliminating the need for specialized model designs. To improve the recommendation performance of LLMs, we introduce a multimodal fusion module for item representation and a sequence-in-set-out approach for efficient candidate generation. When applied to industrial-scale data, our LLM achieves competitive results with expert models elaborately designed for different recommendation tasks. Furthermore, our analysis reveals that recommendation outcomes are highly sensitive to text input, highlighting the potential of prompt engineering in optimizing industrial-scale recommender systems.

Problem

• LLMs as universal recommendation learners
• Multimodal fusion enhances item representation
• Prompt engineering optimizes recommender systems

Innovation

• Multimodal fusion for item representation
• Sequence-in-set-out candidate generation
• Prompt engineering for text optimization