Towards Efficient Whole Page Ranking with One-Pass Inference
Abstract
Whole-page ranking jointly optimizes advertisements and organic content to balance business outcomes and user engagement under complex layout and business constraints. Existing non-autoregressive (NAR) approaches enable efficient page generation but typically score candidate items independently of previously selected items, limiting their ability to capture order-dependent interactions. Autoregressive (AR) methods address this limitation through sequential conditioning, but incur substantial inference overhead that challenges large-scale deployment. Bridging this quality-efficiency gap remains a challenge in industrial whole-page ranking. We propose Multi-Item Scoring (MIS), a prefix-aware NAR-style framework that incorporates sequential context into efficient multi-item scoring. MIS integrates global candidate interactions, localized prefix conditioning, and position-aware scoring within a shared representation architecture. By computing reusable representations once per request and performing lightweight prefix-conditioned scoring during constrained beam search, MIS captures important sequential dependencies without repeated full-model inference. Experiments on KuaiRand-1K and RL4RS demonstrate that MIS approaches the generation quality of an autoregressive CGA baseline while achieving 4.07x and 3.54x faster CPU decoding, respectively. In a randomized evaluation on a large-scale social-video platform, MIS improves the realization metric by 0.104% and watch time by 0.058%, while reducing ranking-service CPU consumption by 9.73%. These results demonstrate that prefix-aware multi-item scoring can narrow the quality gap between NAR and AR generation while retaining the efficiency required for industrial scalability.
Related
- Estimation and Inference for One-pass Streaming Function-on-Function Regression
- Human-Anchored Inference for Ranking New Models with Large Language Model Judges
- RLVR is a Kernel, Not a Function: Statistical Inference for pass@k Crossovers
- Whole-brain modeling of dynamic causal circuits in human cognition using amortized variational inference
- Semiparametrically efficient semi-supervised statistical inference
BibTeX
@misc{xu2026towards,
title = {{Towards Efficient Whole Page Ranking with One-Pass Inference}},
author = {Qi Xu and Zhigang Hua and Zijie Lei and Yan Xie and Jizhe Zhang and Lynn Lu and Delun Du and Sha Hua and Ren Mao and Ke Pan and Haomin Yu and Sandeep Pandey and Luke Simon},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.05542},
eprint = {2611.05542},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.05542},
}
RIS
TY - UNPB
TI - Towards Efficient Whole Page Ranking with One-Pass Inference
AU - Qi Xu
AU - Zhigang Hua
AU - Zijie Lei
AU - Yan Xie
AU - Jizhe Zhang
AU - Lynn Lu
AU - Delun Du
AU - Sha Hua
AU - Ren Mao
AU - Ke Pan
AU - Haomin Yu
AU - Sandeep Pandey
AU - Luke Simon
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.05542
UR - https://arxiv.org/abs/2611.05542
AB - Whole-page ranking jointly optimizes advertisements and organic content to balance business outcomes and user engagement under complex layout and business constraints. Existing non-autoregressive (NAR) approaches enable efficient page generation but typically score candidate items independently of previously selected items, limiting their ability to capture order-dependent interactions. Autoregressive (AR) methods address this limitation through sequential conditioning, but incur substantial inference overhead that challenges large-scale deployment. Bridging this quality-efficiency gap remains a challenge in industrial whole-page ranking. We propose Multi-Item Scoring (MIS), a prefix-aware NAR-style framework that incorporates sequential context into efficient multi-item scoring. MIS integrates global candidate interactions, localized prefix conditioning, and position-aware scoring within a shared representation architecture. By computing reusable representations once per request and performing lightweight prefix-conditioned scoring during constrained beam search, MIS captures important sequential dependencies without repeated full-model inference. Experiments on KuaiRand-1K and RL4RS demonstrate that MIS approaches the generation quality of an autoregressive CGA baseline while achieving 4.07x and 3.54x faster CPU decoding, respectively. In a randomized evaluation on a large-scale social-video platform, MIS improves the realization metric by 0.104% and watch time by 0.058%, while reducing ranking-service CPU consumption by 9.73%. These results demonstrate that prefix-aware multi-item scoring can narrow the quality gap between NAR and AR generation while retaining the efficiency required for industrial scalability.
ER -