Multimodal Advances in Recommendation Systems

The field of recommendation systems is witnessing a significant shift towards incorporating multimodal data and techniques to improve performance and address long-standing challenges. Researchers are exploring the potential of multimodal learning to combine collaborative and side information, narrowing the modality gap and providing accurate recommendations even in missing modality scenarios. Notable papers in this area include:

  • Understanding Embedding Scaling in Collaborative Filtering, which discovers two novel phenomena when scaling collaborative filtering models and provides a theoretical analysis of the noise robustness of these models.
  • Multimodal-enhanced Federated Recommendation, which proposes a novel multimodal fusion mechanism in federated recommendation settings, enabling fine-grained knowledge sharing among similar users while retaining individual preferences.

Sources

Understanding Embedding Scaling in Collaborative Filtering

Optimizing Product Deduplication in E-Commerce with Multimodal Embeddings

Harnessing Multimodal Large Language Models for Personalized Product Search with Query-aware Refinement

Robust Denoising Neural Reranker for Recommender Systems

Single-Branch Network Architectures to Close the Modality Gap in Multimodal Recommendation

Multimodal-enhanced Federated Recommendation: A Group-wise Fusion Approach

Multimodal Representation-disentangled Information Bottleneck for Multimodal Recommendation

Built with on top of