{"slug": "towards-efficient-whole-page-ranking-with-one-pass-inference", "title": "Towards Efficient Whole Page Ranking with One-Pass Inference", "summary": "A new arXiv paper (2611.05542v1) proposes Multi-Item Scoring (MIS), a prefix-aware non-autoregressive framework for whole-page ranking that approaches the generation quality of an autoregressive CGA baseline while decoding 4.07x and 3.54x faster on CPU on the KuaiRand-1K and RL4RS benchmarks, respectively. In a randomized evaluation on a large-scale social-video platform, MIS improved the realization metric by 0.104% and watch time by 0.058% while cutting ranking-service CPU consumption by 9.73%. The authors present MIS as a way to narrow the quality-efficiency gap between non-autoregressive and autoregressive page generation for industrial deployment.", "body_md": "arXiv:2611.05542v1 Announce Type: new \nAbstract: 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.\n  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.\n  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.", "url": "https://wpnews.pro/news/towards-efficient-whole-page-ranking-with-one-pass-inference", "canonical_source": "https://arxiv.org/abs/2611.05542", "published_at": "2026-11-05 05:00:00+00:00", "updated_at": "2026-11-05 05:21:36.635920+00:00", "lang": "en", "topics": ["machine-learning", "ai-research", "ai-infrastructure"], "entities": ["Multi-Item Scoring", "KuaiRand-1K", "RL4RS", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/towards-efficient-whole-page-ranking-with-one-pass-inference", "markdown": "https://wpnews.pro/news/towards-efficient-whole-page-ranking-with-one-pass-inference.md", "text": "https://wpnews.pro/news/towards-efficient-whole-page-ranking-with-one-pass-inference.txt", "jsonld": "https://wpnews.pro/news/towards-efficient-whole-page-ranking-with-one-pass-inference.jsonld"}}