Uncovering Cross-Domain Recommendation Ability of Large Language Models

TL;DR

The work addresses cross-domain recommendation when the target domain lacks historical data by using large language models to transfer knowledge from a related source domain. The authors propose LLM4CDR, a prompt-based pipeline that builds source-domain histories, generates a constrained candidate list, and uses domain-aware guidance to steer LLMs in cross-domain re-ranking. Key findings show that small domain gaps and larger LLMs yield the strongest gains (e.g., MAP improvements on the order of tens of percent), while domain-specific guidance further enhances performance; the approach is most effective with a single closely related source domain. These results suggest practical potential for LLM-driven CDR in low-resource domains, especially when careful input selection, prompt design, and guidance are employed. The work paves the way for future research into robust CDR prompts, data preprocessing strategies, and scalable evaluation across more domains and LLM architectures.

Abstract

Cross-Domain Recommendation (CDR) seeks to enhance item retrieval in low-resource domains by transferring knowledge from high-resource domains. While recent advancements in Large Language Models (LLMs) have demonstrated their potential in Recommender Systems (RS), their ability to effectively transfer domain knowledge for improved recommendations remains underexplored. To bridge this gap, we propose LLM4CDR, a novel CDR pipeline that constructs context-aware prompts by leveraging users' purchase history sequences from a source domain along with shared features between source and target domains. Through extensive experiments, we show that LLM4CDR achieves strong performance, particularly when using LLMs with large parameter sizes and when the source and target domains exhibit smaller domain gaps. For instance, incorporating CD and Vinyl purchase history for recommendations in Movies and TV yields a 64.28 percent MAP 1 improvement. We further investigate key factors including source domain data, domain gap, prompt design, and LLM size, which impact LLM4CDR's effectiveness in CDR tasks. Our results highlight that LLM4CDR excels when leveraging a single, closely related source domain and benefits significantly from larger LLMs. These insights pave the way for future research on LLM-driven cross-domain recommendations.

Figures (4)

• Figure 1: The LLM4CDR framework consists of information preprocessing, prompt generation, and recommendation results generation. The information preprocessing involves generating source domain history, creating a candidate list, and preparing recommendation guidelines. The information obtained from the preprocessing stage is highlighted in bold in the prompt generation section. The prompt generation stage involves task domain adaptation, conditional information, recommendation guidance, and task description (each marked by different colors).
• Figure 2: User purchase histories with different history lengths.
• Figure 3: Recommendation performance of LLM4CDR with different source domain history length.
• Figure 4: Performance comparison with/without recommendation guidance.