发表机构
Tianjin University; Meituan; Institute of Software Chinese Academy of Sciences（天津大学; 美团; 中国科学院软件研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对工业多模态推荐系统的存储与延迟开销问题，提出SA-RSQ稀疏表示框架，经实验验证其在重构性能与CTR间的权衡表现良好，在线测试实现CTR与CPM的相对提升。
AI 中文摘要
在工业推荐系统中部署高维多模态特征会产生大量存储与延迟开销。硬量化虽紧凑但会引入边界失真，而密集软量化则使表示质量受限于有限的存储预算。我们提出了基于稀疏激活的残差软量化（SA-RSQ），其采用Top-K稀疏路由与softmax权重来存储紧凑的（索引、概率）元组。所存储的元组将每个物品的存储与码本维度解耦；在固定选定支持度的情况下，梯度可通过路由权重与加权重构传播，无需依赖直通估计器。在专有外卖广告数据集上的实验表明，在每个物品8-48字节的存储预算范围内，该方法在重构性能与CTR之间取得了良好的权衡。初步的下一个分布预测研究及为期一周的在线A/B测试进一步证明了SA-RSQ的实用潜力，其CTR相对提升2.51%，CPM相对提升3.66%。
英文摘要
Deploying high-dimensional multimodal features in industrial recommender systems incurs substantial storage and latency overhead. Hard quantization is compact but introduces boundary distortion, whereas dense soft quantization couples representation quality to the limited storage budget. We propose Sparse Activation-based Residual Soft Quantization (SA-RSQ), which uses Top-K sparse routing and softmax weights to store compact (Index, Probability) tuples. The stored tuples decouple per-item storage from codebook dimensionality; for a fixed selected support, gradients propagate through the routing weights and weighted reconstruction without relying on a straight-through estimator. Experiments on a proprietary food-delivery advertising dataset show favorable reconstruction-performance and CTR trade-offs across storage budgets of 8-48 bytes per item. A preliminary Next-Distribution Prediction study and a one-week online A/B test further demonstrate the practical potential of SA-RSQ, with relative lifts of +2.51% in CTR and +3.66% in CPM.
