UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866918532621533184 |
|---|---|
| author | Li, Hanyu Hsu, Yi-Ping Mantha, Aditya Agarwal, Prabhat Bhasin, Laksh Wang, Jialu Lin, Hongtao Huang, Bella Li, Yaxin Li, Xinyi Wang, Chuxi Rajesh, Kousik Razaghi, Hooshmand Shokri Li, Shunyao Qin, Zongyue Yang, Jaewon Li, James Badani, Dhruvil Deven Xu, Jiajing Rosenberg, Charles |
| author_facet | Li, Hanyu Hsu, Yi-Ping Mantha, Aditya Agarwal, Prabhat Bhasin, Laksh Wang, Jialu Lin, Hongtao Huang, Bella Li, Yaxin Li, Xinyi Wang, Chuxi Rajesh, Kousik Razaghi, Hooshmand Shokri Li, Shunyao Qin, Zongyue Yang, Jaewon Li, James Badani, Dhruvil Deven Xu, Jiajing Rosenberg, Charles |
| contents | Modern recommendation systems predominantly train retrieval and ranking as separate models despite both increasingly relying on large transformers encoding the same user behavior data, duplicating parameters, compute, and serving cost. Prior work unifies the model architecture but not the full pipeline: input formats, training procedures, and serving stacks remain fragmented across stages. We present UniPinRec, which achieves full-stack unification of retrieval and ranking at Pinterest: one input format, one model, one training stage, deployed within existing serving infrastructure. A shared transformer encodes the user action sequence into candidate-independent representations that branch into retrieval (ANN dot-product) and ranking (cross-attention) via task-specific heads. Three ideas make this work: (1) Masked Action Modeling (MAM) eliminates interleaving, enabling weight sharing without doubling context length; (2) Blended training examples pair action sequences with feedview impression slates to satisfy both objectives jointly; (3) Cross-stage KV cache sharing reuses user-history computation from retrieval for ranking, reducing total FLOPs versus serving two independent models. Deployed in the Pinterest core surfaces, UniPinRec delivers approximately +1% online engagement lift while cutting end-to-end serving latency by 11.1% and lifting QPS by 63.6%. To our knowledge, this is the first full-stack unification of retrieval and ranking, covering inputs, model, training and serving, deployed in a production recommendation system. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2606_00422 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale Li, Hanyu Hsu, Yi-Ping Mantha, Aditya Agarwal, Prabhat Bhasin, Laksh Wang, Jialu Lin, Hongtao Huang, Bella Li, Yaxin Li, Xinyi Wang, Chuxi Rajesh, Kousik Razaghi, Hooshmand Shokri Li, Shunyao Qin, Zongyue Yang, Jaewon Li, James Badani, Dhruvil Deven Xu, Jiajing Rosenberg, Charles Information Retrieval Machine Learning Modern recommendation systems predominantly train retrieval and ranking as separate models despite both increasingly relying on large transformers encoding the same user behavior data, duplicating parameters, compute, and serving cost. Prior work unifies the model architecture but not the full pipeline: input formats, training procedures, and serving stacks remain fragmented across stages. We present UniPinRec, which achieves full-stack unification of retrieval and ranking at Pinterest: one input format, one model, one training stage, deployed within existing serving infrastructure. A shared transformer encodes the user action sequence into candidate-independent representations that branch into retrieval (ANN dot-product) and ranking (cross-attention) via task-specific heads. Three ideas make this work: (1) Masked Action Modeling (MAM) eliminates interleaving, enabling weight sharing without doubling context length; (2) Blended training examples pair action sequences with feedview impression slates to satisfy both objectives jointly; (3) Cross-stage KV cache sharing reuses user-history computation from retrieval for ranking, reducing total FLOPs versus serving two independent models. Deployed in the Pinterest core surfaces, UniPinRec delivers approximately +1% online engagement lift while cutting end-to-end serving latency by 11.1% and lifting QPS by 63.6%. To our knowledge, this is the first full-stack unification of retrieval and ranking, covering inputs, model, training and serving, deployed in a production recommendation system. |
| title | UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale |
| topic | Information Retrieval Machine Learning |
| url | https://arxiv.org/abs/2606.00422 |