QDAR: Query-Guided Distribution-Aware Recommendation for Efficient Intent-Centric Modeling

DOI

10.2991/978-94-6239-799-6_14How to use a DOI?

Keywords

Sequential Recommendation; Efficient AI; Metric Learning; Query-Guided Selection

Abstract

Recent work in sequential recommendation has focused on Transformer-based architectures for their ability to capture long-range dependencies. However, these models face two critical bottlenecks in real-world deployment: quadratic computational complexity O(n²) and an inability to explicitly filter session-irrelevant noise in high-volatility user histories. In this paper, we pro-pose the Query-Guided Distribution-Aware Recommender (QDAR), a lightweight architecture designed for high-precision, low-latency recommendation. QDAR introduces a Query-Guided Interaction Selector that prunes interaction histories in linear O(n) time, focusing exclusively on signals relevant to the current user intent and improving representation robustness. We further replace standard dot-product similarity with a Distribution-Aware Similarity (DAS) metric that accounts for feature covariance via a learnable diagonal Mahalanobis distance. Empirical evaluations on three benchmarks (MovieLens-1M, RetailRocket, and Amazon Beauty) with multi-seed evaluation (seeds 42, 43, 44) and 95% confidence intervals show that QDAR consistently reduces inference latency across all datasets, while achieving its strongest ranking improvements on sparse e-commerce clickstream data (RetailRocket), where it substantially out-performs Transformer baselines across all ranking metrics. A detailed component-wise ablation study confirms the contribution of each architectural module.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Abhishek Abhishek
AU  - P. S. Bishnu
AU  - V. Bhattacharjee
PY  - 2026
DA  - 2026/09/30
TI  - QDAR: Query-Guided Distribution-Aware Recommendation for Efficient Intent-Centric Modeling
BT  - Proceedings of the International Conference on Emerging Trends and Technologies in Applications of Computer Science (ICETTACS 2026)
PB  - Atlantis Press
SP  - 189
EP  - 201
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-799-6_14
DO  - 10.2991/978-94-6239-799-6_14
ID  - Abhishek2026
ER  -