用大模型生成伪标签，解决推荐系统预排序的曝光偏差问题。
Generative Pseudo-Labeling for Pre-Ranking with LLMs
- 用大模型生成内容感知的无偏伪标签，匹配用户兴趣锚点。
- 上线后点击率提升3.07%，长尾内容发现率显著提高。
- 适合大规模推荐系统优化，尤其关注长尾内容的场景。
预排序是工业级推荐系统的关键阶段，需高效评分数千个召回候选项。主要挑战在于训练与服务之间的差异：预排序模型仅在曝光交互上训练，却需在线服务中对所有召回候选项（包括未曝光项）进行评分。这种不匹配导致严重样本选择偏差，削弱泛化能力，尤其影响长尾内容。现有去偏方法多依赖启发式策略（如负采样）或从有偏排序器蒸馏，要么错误将可能的未曝光项标记为负样本，要么将曝光偏差传播至伪标签。本文提出生成式伪标签（GPL），利用大语言模型（LLMs）为未曝光项生成无偏、内容感知的伪标签，显式对齐训练分布与线上服务空间。通过离线生成用户特定的兴趣锚点，并在冻结语义空间中匹配候选项，GPL 提供高质量监督且不增加在线延迟。部署于大规模生产系统后，点击率提升3.07%，显著增强推荐多样性与长尾内容发现能力。
原文摘要 · Abstract (English)
Pre-ranking is a critical stage in industrial recommendation systems, tasked with efficiently scoring thousands of recalled items for downstream ranking. A key challenge is the train-serving discrepancy: pre-ranking models are trained only on exposed interactions, yet must score all recalled candidates -- including unexposed items -- during online serving. This mismatch not only induces severe sample selection bias but also degrades generalization, especially for long-tail content. Existing debiasing approaches typically rely on heuristics (e.g., negative sampling) or distillation from biased rankers, which either mislabel plausible unexposed items as negatives or propagate exposure bias into pseudo-labels. In this work, we propose Generative Pseudo-Labeling (GPL), a framework that leverages large language models (LLMs) to generate unbiased, content-aware pseudo-labels for unexposed items, explicitly aligning the training distribution with the online serving space. By offline generating user-specific interest anchors and matching them with candidates in a frozen semantic space, GPL provides high-quality supervision without adding online latency. Deployed in a large-scale production system, GPL improves click-through rate by 3.07%, while significantly enhancing recommendation diversity and long-tail item discovery.
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。