Reproduce “The Best of the Two Worlds: Harmonizing Semantic and Hash IDs for Sequential Recommendation” 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
Ziwei Liu, Yejing Wang, Wanyu Wang, Wang Zejian, Qidong Liu, Zijian Zhang, Chong Chen, Wei Huang and 1 more
A framework named H$^2$Rec integrates Hash IDs and Semantic IDs in sequential recommender systems through a dual-branch architecture and dual-level alignment, effectively resolving the trade-off between head and tail item recommendation performance and yielding gains up to 11.88% offline and 0.89% in online A/B tests.
Key insights
- The paper formally identifies and characterizes 'Collaborative Overwhelming' in SID-based methods, clarifying why direct SID application can degrade performance for popular items.
- Harmonizing unique collaborative identifiers (HIDs) with multi-granular semantic representations (SIDs) through selective knowledge transfer can overcome the inherent head-tail item recommendation trade-off.
- Adaptive multi-granularity semantic modeling and explicit alignment between semantic and collaborative spaces are crucial for robust sequential recommendation across diverse item popularity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.