MARS: Multi-resolution Adaptive Routing for Sequential Recommendation

arXiv:2610.07505v1 Announce Type: new 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 extit{temporal aliasing}. We propose extbf{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 1.02imes that baseline's warm-cache serving latency for 1{,}000 candidates per user.

Related

• BAHSD: Bridging the Long-tail Gap via Adaptive Distillation in Black-box Sequential Recommendation
• TimeRoute: Time-Aware Modality Routing and Diffusion for Multi-Modal Recommendation
• From Feature Interaction to Feature Transport - A Unified Block for Scalable Recommendation Models

Source: arXiv cs.AI | 2026-10-07