Reproduce “Item Representations Matter: Bridging Hard Semantic IDs and Soft Prototype Mixtures for Generative Retrieval” 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
Vicente Balmaseda, Dongqi Fu, Jiarui Feng, Ren chen, Hanqing Zeng, Tianxin Wei, Yinglong Xia, Tianbao Yang and 1 more
CAP-Mix represents items as soft blends of shared prototypes plus trainable item-specific vectors, and trains these components jointly with a flow recommender. In an offline semantic-ID comparison on Industrial, it scores 3.47 HR@5 versus 3.02 for a hard table. Across six benchmarks, it leads HR@5 on all six; evidence is offline, not online.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.