Search papers, labs, and topics across Lattice.
The paper introduces Hierarchical Sparse Activation Compression (HiSAC), a novel framework for efficient personalized sequence modeling in recommender systems that addresses limitations in identifying user-specific interest centers and accurately assigning behaviors. HiSAC encodes interactions into multi-level semantic IDs, constructs a global hierarchical codebook, and uses a hierarchical voting mechanism to sparsely activate personalized interest-agents. Experimental results from online A/B tests on Taobao's "Guess What You Like" homepage demonstrate a 1.65% CTR uplift, highlighting HiSAC's scalability and effectiveness in capturing long-tail preferences while achieving significant compression and cost reduction.
Taobao's recommender system just got a 1.65% CTR boost by compressing ultra-long user behavior sequences with a hierarchical codebook and sparse attention, proving that personalized interest centers can be learned efficiently.
Modern recommender systems leverage ultra-long user behavior sequences to capture dynamic preferences, but end-to-end modeling is infeasible in production due to latency and memory constraints. While summarizing history via interest centers offers a practical alternative, existing methods struggle to (1) identify user-specific centers at appropriate granularity and (2) accurately assign behaviors, leading to quantization errors and loss of long-tail preferences. To alleviate these issues, we propose Hierarchical Sparse Activation Compression (HiSAC), an efficient framework for personalized sequence modeling. HiSAC encodes interactions into multi-level semantic IDs and constructs a global hierarchical codebook. A hierarchical voting mechanism sparsely activates personalized interest-agents as fine-grained preference centers. Guided by these agents, Soft-Routing Attention aggregates historical signals in semantic space, weighting by similarity to minimize quantization error and retain long-tail behaviors. Deployed on Taobao's"Guess What You Like"homepage, HiSAC achieves significant compression and cost reduction, with online A/B tests showing a consistent 1.65% CTR uplift -- demonstrating its scalability and real-world effectiveness.
DAMO