Reproduce “Decomposing Staleness in Recommender Systems: A Dual-Filter Framework for Supersession and Decay” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Di Bai, Feng Han, Zhenwei Tang, Jintao Liu, Luoshu Wang, Jialu Liu

Researchers at Google LLC introduced SDF, a dual-filter framework that proactively addresses content staleness in recommender systems by decomposing it into supersession and intrinsic decay. The deployed system in Google Discover reduced user-filed staleness reports by 54.9% and achieved significant savings in serving computational resources.

Key insights

  • Decomposing content staleness into distinct mechanisms (supersession and intrinsic decay) enables more targeted and effective mitigation strategies compared to monolithic approaches.
  • Large Language Models can effectively serve as teachers for generating high-quality, rationale-augmented synthetic training data, facilitating the development of scalable student models for complex relational judgments.
  • Proactive filtering of stale content upstream in the recommendation pipeline not only improves user experience but also yields substantial computational efficiency gains by reducing downstream processing of irrelevant items.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.