English(EN) From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs

新的ReST框架扩展Transformer用于工业推荐排序

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-02 04:00

一篇新的研究论文介绍ReST，一个旨在扩展Transformer模型用于工业推荐排序系统的框架。ReST通过结合双门控注意力、旋转位置和时间嵌入以及稳定的残差归一化，解决了嘈杂和不规则的行为序列等挑战。为了处理计算不对称性，它将排序过程分解为一个可重用的编码器和一个轻量级的交叉解码器，从而能够在严格的延迟预算内实现高效服务。在广告平台上进行的A/B测试表明，ReST在50毫秒P99预算内将AUC提高了1.31%，收入提高了11.93%，从而实现了全面生产部署。 AI

影响该框架可以显著提高生产环境中推荐系统的效率和准确性。

排序理由该集群包含一篇详细介绍新模型架构及其性能的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 31 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍新模型架构及其性能的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Jie Chen, Xiangqian Yu, Yanchao Lian, Tan Lu, Run Yang, Zhengchun Shang, Xing Wang, Cheng Chen, Ke Hu, Qiang Li, Tianjiu Yin, Xiaobing Liu · 2026-09-02 04:00

从语言到行为：使用原生推荐设计扩展序列Transformer用于工业推荐排序

arXiv:2609.01240v1 Announce Type: cross Abstract: Scaling Transformers has driven large gains in language modeling, but transplanting this to behavior-sequence modeling in production ranking is challenging: recommendation differs in signal quality, where behavior sequences are no…

本文如何被排名

报道来源 [1]

从语言到行为：使用原生推荐设计扩展序列Transformer用于工业推荐排序

相关实体

相关话题