The scaling laws for recommender systems have been increasingly validated, where MetaFormer-based architectures consistently benefit from increased model depth, hidden dimensionality, and user behavior sequence length. However, whether representation capacity scales proportionally with parameter growth remains unexplored. Prior studies on RankMixer reveal that the effective rank of token representations exhibits a damped oscillatory trajectory across layers, failing to increase consistently with depth and even degrading in deeper layers. Motivated by this observation, we propose RankUp, an architecture designed to mitigate representation collapse and enhance expressive capacity through randomized permutation splitting over sparse features, a multi-embedding paradigm, global token integration and crossed pretrained embedding tokens. RankUp has been fully deployed in large-scale production across Weixin Video Accounts, Official Accounts and Moments, yielding GMV improvements of 3.41%, 4.81% and 2.12%, respectively.
@article{arxiv.2604.17878,
title = {RankUp: Towards High-rank Representations for Large Scale Advertising Recommender Systems},
author = {Jin Chen and Shangyu Zhang and Bin Hu and Chao Zhou and Junwei Pan and Gengsheng Xue and Wentao Ning and Gengyu Weng and Wang Zheng and Shaohua Liu and Zeen Xu and Chengyuan Mai and Shijie Quan and Tingyu Jiang and Lifeng Wang and Shudong Huang and Chengguo Yin and Haijie Gu and Jie Jiang},
journal= {arXiv preprint arXiv:2604.17878},
year = {2026}
}