End-to-End Semantic ID Generation for Generative Advertisement Recommendation

Researchers from Alibaba and other institutions introduced UniSID, a new framework for generating Semantic IDs (SIDs) in generative recommendation systems. It addresses key flaws in current two-stage methods by jointly optimizing embeddings and IDs end-to-end. This approach prevents semantic degradation and error accumulation. UniSID outperforms state-of-the-art methods, achieving up to a 4.62% improvement in Hit Rate metrics for downstream advertising recommendation tasks.