Multimodal Decoupling for Diversified Recommendation through Semantic and Structural Shaping

Key points

• Recommender systems often suffer from over-concentrated results when optimized for accuracy, limiting recommendation diversity.1
• On the one hand, multimodal recommender systems provide rich semantic signals, yet are often treated as auxiliary to ID embeddings, leading to ineffective semantic utilization for diversity.1
• Specifically, we first introduce an MLLM-based Decoupled Feature Extraction module to disentangle multimodal content into informative representations.1
• To address structural limitations, we propose two complementary modules: Alternating Dual-Graph Propagation, which alleviates representation collapse by integrating collaborative and multimodal propagation, and Entropy-based Diversity Regularization, which mitigates optimization bias by promoting balanced category exposure.1

Sources (1)

• [1]Multimodal Decoupling for Diversified Recommendation through Semantic and Structural Shaping
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 08:46 AM

  To address these challenges, we propose MOSS, a unified framework for multimodal decoupling through semantic and structural shaping.

  Recommender systems often suffer from over-concentrated results when optimized for accuracy, limiting recommendation diversity.

Extractive summary: sentences quoted from the sources.