Salvato in:
| Autori principali: | , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2506.00450 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916770504245248 |
|---|---|
| author | Lyu, Wenhan Tyagi, Devashish Yang, Yihang Li, Ziwei Somani, Ajay Shanmugasundaram, Karthikeyan Andrejevic, Nikola Adeputra, Ferdi Zeng, Curtis Singh, Arun K. Ransan, Maxime Jain, Sagar |
| author_facet | Lyu, Wenhan Tyagi, Devashish Yang, Yihang Li, Ziwei Somani, Ajay Shanmugasundaram, Karthikeyan Andrejevic, Nikola Adeputra, Ferdi Zeng, Curtis Singh, Arun K. Ransan, Maxime Jain, Sagar |
| contents | Long user history is highly valuable signal for recommendation systems, but effectively incorporating it often comes with high cost in terms of data center power consumption and GPU. In this work, we chose offline embedding over end-to-end sequence length optimization methods to enable extremely long user sequence modeling as a cost-effective solution, and propose a new user embedding learning strategy, multi-slicing and summarization, that generates highly generalizable user representation of user's long-term stable interest. History length we encoded in this embedding is up to 70,000 and on average 40,000. This embedding, named as DV365, is proven highly incremental on top of advanced attentive user sequence models deployed in Instagram. Produced by a single upstream foundational model, it is launched in 15 different models across Instagram and Threads with significant impact, and has been production battle-proven for >1 year since our first launch. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_00450 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DV365: Extremely Long User History Modeling at Instagram Lyu, Wenhan Tyagi, Devashish Yang, Yihang Li, Ziwei Somani, Ajay Shanmugasundaram, Karthikeyan Andrejevic, Nikola Adeputra, Ferdi Zeng, Curtis Singh, Arun K. Ransan, Maxime Jain, Sagar Information Retrieval Machine Learning Long user history is highly valuable signal for recommendation systems, but effectively incorporating it often comes with high cost in terms of data center power consumption and GPU. In this work, we chose offline embedding over end-to-end sequence length optimization methods to enable extremely long user sequence modeling as a cost-effective solution, and propose a new user embedding learning strategy, multi-slicing and summarization, that generates highly generalizable user representation of user's long-term stable interest. History length we encoded in this embedding is up to 70,000 and on average 40,000. This embedding, named as DV365, is proven highly incremental on top of advanced attentive user sequence models deployed in Instagram. Produced by a single upstream foundational model, it is launched in 15 different models across Instagram and Threads with significant impact, and has been production battle-proven for >1 year since our first launch. |
| title | DV365: Extremely Long User History Modeling at Instagram |
| topic | Information Retrieval Machine Learning |
| url | https://arxiv.org/abs/2506.00450 |