LENS: A Staged Design for Interaction Granularity in Sequential CTR Prediction CCS: Information systems Recommender systems CCS: Information systems Retrieval models and ranking
Yuan Wang, Yue Liu, Jun Zhang, Jie Jiang · 2026
Read LENS: A Staged Design for Interaction Granularity in Sequential CTR Prediction CCS: Information systems Recommender systems CCS: Information systems Retrieval models and ranking by Yuan Wang, Yue Liu, Jun Zhang, Jie Jiang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-25583.txt.
Description
Wang and colleagues ask how finely a target item should interact with a user's past behavior in click prediction. Raw-item models score every past item directly but falter on rare items. Latent-query models blend the target into a coarser query and dilute its control. Their LENS design restores that control with a query gate and a position bias that both take the target as a condition. It adds a static position reference called QueryPos, and every backbone-dataset combination shows a positive total gain.
Catalog record
| CREATOR | Yuan Wang, Yue Liu, Jun Zhang, Jie Jiang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71506 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.25583