Utility-Aware Generative Recommendation: Utility-Enriched Semantic Codes and Utility-Guided Decoding

📄 arXiv: 2611.02981v1 📥 PDF

作者: Yang Hu, Duo Wang, Ning Li, Kang Li, Junda She, Chengcheng Guo, Kuo Cai, Bin Liu, Youhua Liu, Xiaoshuang Chen

分类: cs.LG

发布日期: 2026-11-02

💡 一句话要点

提出UAGR以解决生成推荐中的效用不足问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 生成推荐 效用增强 多模态推荐 语义编码 效用引导解码 在线测试 视频推荐

📋 核心要点

  1. 现有生成推荐方法在SID构建中未能充分反映物品效用，导致相似物品难以区分。
  2. 论文提出UAGR，通过效用增强的语义编码和效用引导的解码，提升推荐效果。
  3. 实验表明，UAGR在真实流量下的在线测试中提高了视频观看时间0.206%，显示出良好的应用效果。

📝 摘要（中文）

生成推荐在大规模工业推荐系统中被广泛采用，其中多模态物品嵌入被量化为离线语义标识符（SID），并通过自回归生成其SID来检索物品。然而，传统的SID构建往往无法充分反映物品的效用，导致语义相似但效用不同的物品难以区分。为了解决这些问题，本文提出了UAGR，通过对SID构建和解码与下游多目标效用对齐，提出了效用增强的语义编码和效用引导的解码方法。实验结果表明，UAGR在真实流量下的在线A/B测试中提高了视频观看时间0.206%。

🔬 方法详解

问题定义：本文旨在解决传统生成推荐方法中SID构建不足以反映物品效用的问题，导致高效用物品在生成过程中被错误地修剪。

核心思路：UAGR通过将效用信息融入SID构建和解码过程，确保生成的推荐结果能够更好地反映物品的实际效用。

技术框架：UAGR的整体架构包括效用增强的语义编码模块和效用引导的解码模块，前者负责在SID中嵌入物品效用，后者则在生成过程中考虑效用预测。

关键创新：UAGR的核心创新在于效用增强的语义编码，它在不增加SID长度或词汇量的情况下，支持效用的增量更新，显著提升了推荐的准确性。

关键设计：在设计中，采用了多种后曝光结果的预测来估计候选物品的效用，并将这些估计融入到束搜索的评分中，以减少高效用路径的提前修剪风险。

🖼️ 关键图片

fig_0
fig_1

📊 实验亮点

UAGR在真实流量下的在线A/B测试中，成功提高了视频观看时间0.206%。与传统方法相比，UAGR在效用反映和推荐准确性方面表现出显著提升，展示了其在实际应用中的有效性和可部署性。

🎯 应用场景

该研究的潜在应用领域包括视频推荐、电子商务和社交媒体等场景，能够有效提升用户体验和平台的用户粘性。通过更准确的推荐，平台可以增加用户的观看时间和互动率，进而提高整体收益。未来，该方法有望扩展到更多类型的推荐系统中，推动个性化推荐的发展。

📄 摘要（原文）

Generative recommendation has been widely adopted in large-scale industrial recommendation systems, where multimodal item embeddings are quantized into offline semantic identifiers (SIDs) and items are retrieved by autoregressively generating their SIDs. However, conventional SID construction often underrepresents item utility reflected in post-exposure behaviors such as shares, long views, and video completion, making semantically similar items with different utility insufficiently distinguishable while struggling to accommodate rapidly changing utility signals. Moreover, generation likelihood may be misaligned with downstream utility, causing high-utility candidates to be prematurely pruned during beam search before post-generation reranking or training-time reward alignment can effectively compensate. To address these limitations, we propose UAGR, which aligns both SID construction and decoding with downstream multi-objective utility. Our Utility-Enriched Semantic Codes preserve the original semantic information while incorporating item-level utility without increasing SID length or vocabulary size, supporting incremental utility updates with no additional tokenization or sequence-decoding cost. Meanwhile, our Utility-Guided Decoding estimates candidate utility by combining predictions of multiple post-exposure outcomes and incorporates these estimates into beam scoring, reducing the risk of prematurely pruning high-utility paths while enabling flexible utility control without interfering with generative backbone training. Extensive offline experiments and online A/B tests under real-world traffic demonstrate the effectiveness and deployability of UAGR, which has been fully deployed on Kuaishou and increases video watch time by 0.206\% online.