A Learnable Fully Interacted Two-Tower Model for Pre-Ranking System DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 Conference: Proceedings of the 48th International
Chao Xiong, Xianwen Yu, Wei Xu, Lei Cheng, Chuan Yuan, Linjian Mo · 2025
Read A Learnable Fully Interacted Two-Tower Model for Pre-Ranking System DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 Conference: Proceedings of the 48th International by Chao Xiong, Xianwen Yu, Wei Xu, Lei Cheng, Chuan Yuan, Linjian Mo (2025) 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-2509-12948.txt.
Description
Chao Xiong, Xianwen Yu and four co-authors propose FIT in 2025, a pre-ranking model that lets user and item sides talk early. Pre-ranking picks a short list fast in large shop systems. Classic twin towers score each side alone, which is fast but misses cross links.
FIT adds a learned item store for early cross talk. A light scorer then adds late cross talk between the towers. Speed at test time stays high. Trials on public sets beat top prior pre-rankers by clear gaps. The design shows early talk need not break fast service.
Catalog record
| CREATOR | Chao Xiong, Xianwen Yu, Wei Xu, Lei Cheng, Chuan Yuan, Linjian Mo |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-12948