MARS: Multi-resolution Adaptive Routing for Sequential Recommendation

2026-10-05

AI Summary

This study addresses the "temporal aliasing" problem in long-sequence recommendation, where history compression induces multi-scale semantic loss. To mitigate this, we propose a multi-resolution adaptive routing mechanism that encodes complete user histories via recursive state trajectories with varying half-lives, while employing a sparse routing reader to dynamically select appropriate temporal resolutions for each candidate item, thereby constructing compact, candidate-agnostic user memory representations. Our primary contributions include formally identifying and defining the temporal aliasing failure mode, and designing an efficient multi-resolution memory architecture with sparse routing for long-history modeling. Extensive experiments on three public datasets demonstrate that our approach significantly outperforms strong baselines, with performance gains widening as sequence length increases. Furthermore, the model exhibits strong robustness under behavioral drift scenarios while incurring only marginal inference latency overhead compared to baseline methods.

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.

Problem

• Sequential Recommendation
• Temporal Aliasing
• Long-history Modeling
• Multi-scale Semantics

Innovation

• Sequential Recommendation
• Multi-resolution Memory
• Temporal Aliasing
• Sparse Routing
• Long-history Modeling