Multi resolution routing improves recommendation from long user histories
MARS: Multi-resolution Adaptive Routing for Sequential Recommendation
Artificial Intelligence
Summary
Many recommendation systems try to summarize a user's entire history into a single memory to predict what they might like next. The authors found that user behavior has different time scales—short-term intents, medium-term interests, and long-term preferences—and a single memory mixes these, causing errors called temporal aliasing. They created MARS, a system that keeps multiple memories at different time scales and picks which to use when scoring items. This approach helps recommendations get better, especially when user histories are long or when user behavior changes.
What this means in practice
- •For recommendation system engineers: Build recommendation algorithms that adapt to different time scales of user behavior for improved accuracy with long user histories.
- •For ecommerce platform developers: Improve product recommendations by using multi-resolution memory to handle changing user interests and long browsing histories efficiently.
Authors
Ming Yin, Sixun Dong, Yudong Liu, Wen-Yun Yang, Yunjiang Jiang, Yiran Chen
Abstract
Long-history recommenders often compress each user's history into a compact, candidate-independent memory that is cached and reused to score large candidate pools. We show that real user histories exhibit multi-scale semantic structure, with short-lived intent, medium-term interests, and long-term preferences coexisting in one sequence, and that monolithic cached memories preserve these scales unevenly: linear probes recover recent and mid-range content far worse than long-range content. We call this failure mode \textit{temporal aliasing}. We propose \textbf{MARS}, a multi-resolution user memory that writes the full history into recurrent state tracks anchored to different half-lives, and a sparse routing reader that materializes compact seed memories by selecting the relevant temporal resolutions for each seed, preserving fixed-size candidate scoring. MARS outperforms strong baselines on three public datasets, with gains that grow with history length. Component-matched ablations with paired tests show that temporal diversity and selective routing each contribute beyond what hard-window memories or added capacity provide. The advantage of MARS over its interface-matched baseline also widens after within-user behavioral shifts, at about $1.02\times$ that baseline's warm-cache serving latency for $1{,}000$ candidates per user.