Paper Detail
Hannan Cao, Jun Guo, Haolei Pei, Zhaoqi Zhang, Tianyu Wang, Ziyang Wang, Youchen Sun, Yue Xue, Yucheng Mao, Lintao Yan, Yufei Feng, Shaowei Liu, Rongkun Xing, Feiling Gong, Xinyu Chenli, Cong Xu, Mingge Zhang, Yunjia Zhu, Yajing Zhang, Pengfei Ren, Yue Lin
Industrial recommendation systems typically operate as a \emph{cascade} of retrieval, pre-rank, and fine-rank, but these stages are usually trained and served as separate models, causing repeated user-sequence encoding, isolated optimization, and duplicated engineering effort. Building on OneTrans' model-level unification, we present OneTrans-V2, one Transformer that unifies the entire cascade. It encodes the user behavior sequence once as a shared context while preserving stage-specific candidate features and computation. Joint training lets the three stages reinforce one another and enables in-model knowledge distillation from fine-rank to pre-rank. We scale the shared backbone with sparse mixture-of-experts (MoE), which increases capacity with bounded activated computation, and stabilize scaling with $μ$P-style parameterization. To consolidate objective-specific retrieval channels, we introduce Decision-Conditioned Generative Retrieval (DCGR). DCGR predicts a decision prefix describing the upcoming interaction and generates items conditioned on it, allowing business objectives to steer a single generative process. Finally, Sequence-Native Training (SNT) organizes training around each user's lifelong behavior sequence and amortizes its encoding across exposures. Deployed across all three stages of a large-scale industrial recommendation system, OneTrans-V2 improves gross merchandise value (GMV) by 9.74\% and, with a co-designed serving stack, delivers $3.2\times$ the throughput of the cascade it replaces under the same hardware budget.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@article{cao2026onetrans,
title = {OneTrans-V2: Unifying Retrieval, Pre-rank, and Fine-rank with One Transformer in Industrial Recommender},
author = {Hannan Cao and Jun Guo and Haolei Pei and Zhaoqi Zhang and Tianyu Wang and Ziyang Wang and Youchen Sun and Yue Xue and Yucheng Mao and Lintao Yan and Yufei Feng and Shaowei Liu and Rongkun Xing and Feiling Gong and Xinyu Chenli and Cong Xu and Mingge Zhang and Yunjia Zhu and Yajing Zhang and Pengfei Ren and Yue Lin},
year = {2026},
abstract = {Industrial recommendation systems typically operate as a \textbackslash{}emph\{cascade\} of retrieval, pre-rank, and fine-rank, but these stages are usually trained and served as separate models, causing repeated user-sequence encoding, isolated optimization, and duplicated engineering effort. Building on OneTrans' model-level unification, we present OneTrans-V2, one Transformer that unifies the entire cascade. It encodes the user behavior sequence once as a shared context while preserving stage-specific candida},
url = {https://arxiv.org/abs/2609.28589},
keywords = {cs.IR},
eprint = {2609.28589},
archiveprefix = {arXiv},
}
{}