I don't think popularity in general should be a major concern in recommendations. An item becomes popular only if many users like it.
Rethinking Popularity Bias in Collaborative Filtering via Analytical Vector Decomposition
Finds that popularity bias in collaborative filtering stems from an intrinsic geometric artifact in BPR optimization.
📝 arxiv.org/abs/2512.10688
👨🏽💻 github.com/LingFeng-Liu...
Finds that popularity bias in collaborative filtering stems from an intrinsic geometric artifact in BPR optimization.
📝 arxiv.org/abs/2512.10688
👨🏽💻 github.com/LingFeng-Liu...
Rethinking Popularity Bias in Collaborative Filtering via Analytical Vector Decomposition
Popularity bias fundamentally undermines the personalization capabilities of collaborative filtering (CF) models, causing them to disproportionately recommend popular items while neglecting users' gen...
arxiv.org
December 12, 2025 at 6:35 AM
I don't think popularity in general should be a major concern in recommendations. An item becomes popular only if many users like it.
My student will present two papers at #sigirap2025, both on video moment retrieval. One paper clarifies the research problem and introduces a dataset that reflects real-world video moment search settings, while the other presents a scalable solution.
December 7, 2025 at 10:18 AM
My student will present two papers at #sigirap2025, both on video moment retrieval. One paper clarifies the research problem and introduces a dataset that reflects real-world video moment search settings, while the other presents a scalable solution.