2025/06/22 by Hiba Bederina, Jill-Jênn Vie, Bederina, Hiba +1
Computer Science · #Advanced Graph Neural Networks #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Mobile Crowdsensing and Crowdsourcing #Recommender Systems and Techniques
paper · pdf · doi:10.48550/arxiv.2506.21617
openalex publication_date 2025/06/22 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The challenge of balancing user relevance and content diversity in recommender systems is increasingly critical amid growing concerns about content homogeneity and reduced user engagement. In this work, we propose a novel framework that leverages a multi-objective, contextual sequential sampling strategy. Item selection is guided by Bayesian updates that dynamically adjust scores to optimize diversity. The reward formulation integrates multiple diversity metrics-including the log-determinant volume of a tuned similarity submatrix and ridge leverage scores-along with a diversity gain uncertainty term to address the exploration-exploitation trade-off. Both intra- and inter-batch diversity are modeled to promote serendipity and minimize redundancy. A dominance-based ranking procedure identifies Pareto-optimal item sets, enabling adaptive and balanced selections at each iteration. Experiments on a real-world dataset show that our approach significantly improves diversity without sacrificing relevance, demonstrating its potential to enhance user experience in large-scale recommendation settings.