LLM-I2I: Boost Your Small Item2Item Recommendation Model with Large Language Model

TL;DR

This work tackles data sparsity and noise in item-to-item recommendation by introducing LLM-I2I, a data-centric framework that jointly trains an LLM-based data generator to synthesize user-item interactions and an LLM-based discriminator to filter these interactions before fusing them with real data. A long-tail aware loss with weights $\alpha$ and $\beta$ biases generation toward underrepresented items, improving long-tail coverage. Empirical results on ARD and AEDS show consistent offline gains across diverse I2I backbones, and a production deployment on AliExpress yields RN +6.02% and GMV +1.22% without added latency. The combination of data generation, filtering, and strategic data augmentation demonstrates the practical potential of LLMs for enhancing data-centric recommendation systems while preserving existing architectures.

Abstract

Item-to-Item (I2I) recommendation models are widely used in real-world systems due to their scalability, real-time capabilities, and high recommendation quality. Research to enhance I2I performance focuses on two directions: 1) model-centric approaches, which adopt deeper architectures but risk increased computational costs and deployment complexity, and 2) data-centric methods, which refine training data without altering models, offering cost-effectiveness but struggling with data sparsity and noise. To address these challenges, we propose LLM-I2I, a data-centric framework leveraging Large Language Models (LLMs) to mitigate data quality issues. LLM-I2I includes (1) an LLM-based generator that synthesizes user-item interactions for long-tail items, alleviating data sparsity, and (2) an LLM-based discriminator that filters noisy interactions from real and synthetic data. The refined data is then fused to train I2I models. Evaluated on industry (AEDS) and academic (ARD) datasets, LLM-I2I consistently improves recommendation accuracy, particularly for long-tail items. Deployed on a large-scale cross-border e-commerce platform, it boosts recall number (RN) by 6.02% and gross merchandise value (GMV) by 1.22% over existing I2I models. This work highlights the potential of LLMs in enhancing data-centric recommendation systems without modifying model architectures.

Figures (5)

• Figure 1: A comparative illustration of model-centric vs. data-centric approaches in recommendation system optimization.
• Figure 2: The framework of our proposed LLM-I2I.
• Figure 3: Supervised Fine-Tuning (SFT) of the LLM-based generator and discriminator.
• Figure 4: Data distribution in ARD Toys And Games dataset.
• Figure 5: Recall@K under different settings for the LLM-based generator and discriminator.