Decoupled Entity Representation Learning for Pinterest Ads Ranking CCS: Information systems Recommender systems
Jie Liu, Yinrui Li, Jiankai Sun, Kungang Li, Han Sun, Sihan Wang, Huasen Wu, Siyuan Gao, Paulo Soares, Nan Li, Zhifang Liu, Haoyang Li, Siping Ji, Ling Leng, Prathibha Deshikachar · 2025
Read Decoupled Entity Representation Learning for Pinterest Ads Ranking CCS: Information systems Recommender systems by Jie Liu, Yinrui Li, Jiankai Sun, Kungang Li, Han Sun, Sihan Wang, Huasen Wu, Siyuan Gao, Paulo Soares, Nan Li, Zhifang Liu, Haoyang Li, Siping Ji, Ling Leng, Prathibha Deshikachar (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-2509-04337.txt.
Description
Jie Liu and fourteen co-authors describe how Pinterest builds reusable user and item profiles for ranking advertisements in a 2025 research paper. Each user and each Pin, the platform's image posts, gets an embedding, a numeric summary of its traits. Upstream models learn those summaries from many data sources with complex designs, then refresh them on a schedule instead of computing them live. Downstream ranking and retrieval systems consume the ready-made features.
The split design lets heavy upstream training stay separate from fast ad decisions. Offline tests and live production metrics both improved across tasks, including click and conversion prediction. The framework now runs inside Pinterest's production ad ranking. Other platforms with huge catalogs could copy the decoupled pattern.
Catalog record
| CREATOR | Jie Liu, Yinrui Li, Jiankai Sun, Kungang Li, Han Sun, Sihan Wang, Huasen Wu, Siyuan Gao, Paulo Soares, Nan Li, Zhifang Liu, Haoyang Li, Siping Ji, Ling Leng, Prathibha Deshikachar |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-04337