![[2504.10507] PinRec: Unified Generative Retrieval for Pinterest Recommender Systems](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2504.10507] PinRec: Unified Generative Retrieval for Pinterest Recommender Systems
Summary
The paper introduces PinRec, a unified generative retrieval model for Pinterest's recommendation systems, enhancing performance across various surfaces while adapting to specific business goals.
Why It Matters
PinRec represents a significant advancement in recommender systems by integrating multiple product surfaces into a single model. This approach not only improves user engagement through better recommendations but also streamlines the operational complexity of managing separate models for different surfaces. As personalization becomes increasingly crucial in digital platforms, innovations like PinRec can set new benchmarks for efficiency and effectiveness in recommendation algorithms.
Key Takeaways
- PinRec is the first unified generative retrieval model for Pinterest, enhancing recommendations across various surfaces.
- The model utilizes a pretraining and finetuning approach to adapt to specific business needs.
- Incorporates outcome conditioned generation to align recommendations with surface-specific goals.
- Demonstrated a +4% increase in search saves, showcasing its effectiveness.
- Addresses the challenge of evolving user interests over time in recommendation systems.
Computer Science > Information Retrieval arXiv:2504.10507 (cs) [Submitted on 9 Apr 2025 (v1), last revised 23 Feb 2026 (this version, v5)] Title:PinRec: Unified Generative Retrieval for Pinterest Recommender Systems Authors:Edoardo Botta, Jaewon Yang, Yi-Ping Hsu, Laksh Bhasin, Prabhat Agarwal, Anirudhan Badrinath, Yilin Chen, Jiajing Xu, Charles Rosenberg View a PDF of the paper titled PinRec: Unified Generative Retrieval for Pinterest Recommender Systems, by Edoardo Botta and 8 other authors View PDF HTML (experimental) Abstract:Generative retrieval methods employ sequential modeling techniques, like transformers, to generate candidate items for recommender systems. These methods have demonstrated promising results in academic benchmarks, surpassing traditional retrieval models such as two tower architectures. However, a key limitation is that current approaches require a separate model for each product surface, as building a unified model that accommodates the different business needs of various surfaces has proven challenging. Furthermore, existing methods often fail to capture the evolution of user interests over a sequence, focusing instead on only predicting the next item. This paper introduces PinRec, a novel unified generative retrieval model for all of Pinterest recommendation surfaces, including home feed, search, and related pins. PinRec is pretrained on user activity sequences aggregated across surfaces, then finetuned for each surface using impression data fr...
