HiSAC: Hierarchical Sparse Activation Compression for Ultra-long Sequence Modeling in Recommenders

TL;DR

HiSAC is a hierarchical sparse activation framework that compresses ultra-long user sequences in recommender systems, improving efficiency and accuracy while capturing long-tail preferences, demonstrated by significant online performance gains.

Contribution

It introduces a hierarchical codebook and voting mechanism for personalized interest centers, reducing quantization errors and enhancing long-tail behavior modeling in sequence compression.

Findings

Achieves significant compression and cost reduction in production.

Online A/B tests show a 1.65% CTR uplift.

Effective in capturing long-tail user preferences.

Abstract

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.