Large-scale recommendation systems face "Memory Wall" bottlenecks due to massive, dense embedding tables. While generative retrieval uses discrete tokens for IDs, high-dimensional context still relies on inefficient dense formats. Inspired by computer vision data compression, we propose Dual-purpose Semantic IDs to achieve LLM-level I/O efficiency.
Our methodology uses hierarchical quantization to condense continuous embeddings into discrete Semantic IDs performing two concurrent roles:
- Collaborative Identity: modeling user-item interactions via learnable embedding table;
- Content Reconstruction: using a lightweight Semantic Decoder for on-the-fly embedding approximation.
This approach replaces massive vector storage with on-demand reconstruction, reducing system overhead and data footprints. We demonstrate the efficacy of our framework through offline evaluations and successful online deployment in production-scale ranking and retrieval systems at a major video sharing platform, showing that discrete tokens are indeed all you need for highly efficient, content-rich recommendation.
Blogger's Review: The introduction of Dual-purpose Semantic IDs not only addresses the memory bottleneck issues in traditional recommendation systems but also enhances user experience through efficient data management. This innovative approach deserves further exploration and application in various fields.