MoToRec: Sparse-Regularized Multimodal Tokenization for Cold-Start Recommendation

2026-02-11

AI Summary

This work addresses the challenges of data sparsity and cold-start items in recommender systems by proposing a multimodal recommendation approach based on discrete semantic tokenization. The method introduces a sparsity-regularized Residual Quantized Variational Autoencoder (RQ-VAE) to generate disentangled and interpretable discrete semantic tokens, designs an adaptive rarity-aware augmentation mechanism that prioritizes the optimization of cold-start item representations, and constructs a hierarchical multi-source graph encoder to effectively integrate multimodal collaborative signals. Extensive experiments on three large-scale datasets demonstrate that the proposed approach significantly outperforms state-of-the-art methods in both overall recommendation performance and cold-start scenarios.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Graph neural networks (GNNs) have revolutionized recommender systems by effectively modeling complex user-item interactions, yet data sparsity and the item cold-start problem significantly impair performance, particularly for new items with limited or no interaction history. While multimodal content offers a promising solution, existing methods result in suboptimal representations for new items due to noise and entanglement in sparse data. To address this, we transform multimodal recommendation into discrete semantic tokenization. We present Sparse-Regularized Multimodal Tokenization for Cold-Start Recommendation (MoToRec), a framework centered on a sparsely-regularized Residual Quantized Variational Autoencoder (RQ-VAE) that generates a compositional semantic code of discrete, interpretable tokens, promoting disentangled representations. MoToRec's architecture is enhanced by three synergistic components: (1) a sparsely-regularized RQ-VAE that promotes disentangled representations, (2) a novel adaptive rarity amplification that promotes prioritized learning for cold-start items, and (3) a hierarchical multi-source graph encoder for robust signal fusion with collaborative signals. Extensive experiments on three large-scale datasets demonstrate MoToRec's superiority over state-of-the-art methods in both overall and cold-start scenarios. Our work validates that discrete tokenization provides an effective and scalable alternative for mitigating the long-standing cold-start challenge.

Problem

• cold-start recommendation
• data sparsity
• multimodal content
• item representation
• recommendation systems

Innovation

• discrete tokenization
• cold-start recommendation
• sparse regularization
• multimodal representation
• graph neural networks