TokenMinds : Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems

Qingyun Liu, Bo Yan, Yang Liu, Yuji Roh, Ekansh Sharma, Likang Yin, Emma Olowo, Min-hsuan Tsai, Yuxuan Li, Diego Uribe, Saksham Aggarwal, Siqi Wu, Yuan Hao, Vikas Kedigehalli, Lukasz Heldt, Lichan Hong, Li Wei, Xinyang Yi · 2026

Read TokenMinds : Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems by Qingyun Liu, Bo Yan, Yang Liu, Yuji Roh, Ekansh Sharma, Likang Yin, Emma Olowo, Min-hsuan Tsai, Yuxuan Li, Diego Uribe, Saksham Aggarwal, Siqi Wu, Yuan Hao, Vikas Kedigehalli, Lukasz Heldt, Lichan Hong, Li Wei, Xinyang Yi (2026) 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-2606-25147.txt.

Description

Industrial recommender systems usually model users as dense embedding vectors, whose fixed size constrains what they can express. Qingyun Liu and seventeen colleagues present TokenMinds, a system that gives users discrete tokens as well, extending the PLUM framework from item retrieval to user modelling.

An encoder decoder built from pretrained language models produces both semantic ID based user tokens and dense user embeddings. The dual output keeps the grounded, readable benefits of discrete tokens while staying compatible with downstream models that expect vectors. Because items and users now share one semantic ID vocabulary, long form and short form video behaviour can be modelled in a single system, cutting training and serving costs. TokenMinds is validated through extensive offline experiments and live launches.

Catalog record

CREATORQingyun Liu, Bo Yan, Yang Liu, Yuji Roh, Ekansh Sharma, Likang Yin, Emma Olowo, Min-hsuan Tsai, Yuxuan Li, Diego Uribe, Saksham Aggarwal, Siqi Wu, Yuan Hao, Vikas Kedigehalli, Lukasz Heldt, Lichan Hong, Li Wei, Xinyang Yi
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.25147

Browse the full library · About the Tower of Babel