通过双向门控扩散模型，分离用户行为中的噪声与多层级意图。
GateDiffInt: Gate-Mediated Controllable Diffusion and Multi-Intent LLM Distillation for User Behavior Modeling
- 用可控制的扩散过程增强和去噪用户行为序列
- 从真实转化信号中联合优化去噪与意图提取，提升预测准确率
- 适合大规模推荐系统，尤其在高噪声工业场景中表现突出
现有排序模型仅隐式编码意图，难以解耦不同强度和时间尺度的结构化意图。行为序列中的噪声与意图相互强化，我们称之为噪声-意图耦合（NIC）。噪声会稀释真实意图，而缺乏结构化意图先验使去噪无明确目标。为此，我们提出GateDiffInt，一种用于工业排序的意图交互框架。它利用最终转化信号，联合对齐序列去噪与意图提取。该方法采用可控前向扩散过程配合双重门控机制，以增强并去噪行为序列。随后，大语言模型作为教师，将四种结构化意图——长期、短期、潜在及转化——蒸馏至轻量学生模型。增强后的序列与结构化意图表示通过注意力深度融合，生成面向转化率预测的意图感知表示。在公开数据集与大规模工业数据集上的实验表明，该方法持续优于强基线。在线A/B测试中，服务数亿日活用户，显著提升GMV，已部署至主流量入口，验证了其有效性和生产可用性。
原文摘要 · Abstract (English)
Existing ranking models encode intent only implicitly, making it hard to disentangle structured intents of varying strength and temporal scale. Noise and intent in behavior sequences are mutually reinforcing---we call this Noise--Intent Coupling (NIC). Noise dilutes true intents, while the lack of structured intent priors leaves denoising without a clear target. To address NIC, we propose GateDiffInt, an intent interaction framework for industrial ranking. It uses the final conversion signal to jointly align sequence denoising and intent extraction. GateDiffInt applies a controllable forward diffusion process with dual gating to enhance and denoise behavior sequences. A large language model then acts as teacher to distill four structured intents---long-term, short-term, latent, and conversion into a lightweight student model. The enhanced sequence and structured intent representations are deeply fused via attention to produce intent-aware representations for conversion-rate prediction. Extensive experiments on public and large-scale industrial datasets show consistent gains over strong baselines. In online A/B tests serving hundreds of millions of daily active users, GateDiffInt delivers substantial GMV improvements and has been deployed to primary traffic, confirming both effectiveness and production readiness.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。