LENS: A Staged Design for Interaction Granularityin Sequential CTR Prediction

2026-05-25

AI Summary

This work addresses the challenge of achieving interaction granularity between user behavior sequences and target items in sequential click-through rate (CTR) prediction that simultaneously ensures sparsity robustness and target specificity. To this end, the authors propose the LENs framework, which restores target-specific control within an implicit query architecture through a two-stage design that optimizes query activation and historical retrieval. The framework introduces three key innovations: a Target-Conditioned Query Gate (TCQG), Target-Conditioned Position Bias (TCPB), and Query-Specific Position Bias (QueryPos), enabling fine-grained interactions. Extensive experiments demonstrate consistent performance gains across three mainstream implicit query backbone models and four benchmark datasets, validating the effectiveness and generalization capability of the proposed approach.

Abstract

In sequential CTR prediction, a central design question is at what granularity the target should interact with the user behaviour sequence. Existing models mainly follow two routes. Raw-item architectures such as DIN let the target score each item in the sequence directly. This relies on well-trained item embeddings and becomes brittle for sparse items. Latent-query architectures such as HyFormer, MixFormer, and OneTrans build query representations by combining the target with other information. This is more robust across item-density regimes but blunter: target-specific control is diluted. We propose LENS to restore target-specific control within these coarser bottlenecks. LENS has two modules: a Target-Conditioned Query Gate (TCQG) for query activation and a Target-Conditioned Position Bias (TCPB) for history retrieval. We further introduce Query-Specific Position Bias (QueryPos), a simple static position-aware reference for latent-query backbones. Across three representative latent-query backbones and four datasets, the combined QueryPos+LENS design achieves positive total-gain point estimates in all twelve evaluated backbone--dataset cells. We also identify a density-dependent conditioning rule: as item density decreases, the optimal condition source shifts from item-only to item-plus-sequence.

Problem

• sequential CTR prediction
• interaction granularity
• target-specific control
• item sparsity
• latent-query architectures

Innovation

• sequential CTR prediction
• interaction granularity
• latent-query architecture
• target-conditioned gating
• position bias