Watch time is widely used as a proxy for user satisfaction in video recommendation platforms. However, raw watch times are influenced by confounding factors such as video duration, popularity, and individual user behaviors, potentially distorting preference signals and resulting in biased recommendation models. We propose a novel relative advantage debiasing framework that corrects watch time by comparing it to empirically derived reference distributions conditioned on user and item groups. This approach yields a quantile-based preference signal and introduces a two-stage architecture that explicitly separates distribution estimation from preference learning. Additionally, we present distributional embeddings to efficiently parameterize watch-time quantiles without requiring online sampling or storage of historical data. Both offline and online experiments demonstrate significant improvements in recommendation accuracy and robustness compared to existing baseline methods.

翻译：观看时长在视频推荐平台中被广泛用作用户满意度的代理指标。然而，原始观看时长受到视频时长、流行度及个体用户行为等混杂因素的影响，可能扭曲偏好信号并导致推荐模型产生偏差。本文提出一种新颖的相对优势去偏框架，通过将观看时长与基于用户和物品组别经验推导的参考分布进行比较，实现对其校正。该方法生成基于分位数的偏好信号，并引入两阶段架构，明确将分布估计与偏好学习分离。此外，我们提出分布嵌入方法，无需在线采样或存储历史数据即可高效参数化观看时长分位数。离线和在线实验均表明，相较于现有基线方法，本方法在推荐准确性和鲁棒性方面均有显著提升。