中文

DADF: A Distribution-Aware Debiasing Framework for Watch-Time Regression in Recommender Systems

信息检索 2026-05-19 v1

摘要

Watch-time prediction is a central regression task in short-video recommender systems, where labels are highly long-tailed and residual errors vary systematically across observed watch-time regions. In practice, a model may appear globally calibrated while still overestimating short views and underestimating long views, because opposite errors cancel out in aggregate. Existing methods mainly improve the first-stage watch-time predictor, but often leave such residual distributional bias insufficiently corrected. We propose DADF, a distribution-aware debiasing framework for watch-time regression. Instead of replacing a deployed predictor, DADF performs second-stage multiplicative residual correction on top of it. DADF combines three complementary designs: a dynamic distribution-aware transformation for stabilizing long-tailed correction targets, a debias-factor-aware module for modeling heterogeneous residual patterns using inference-time observable factors, especially video duration, and a multi-label-aware module that exploits auxiliary prediction signals from engagement heads. We evaluate DADF on public short-video benchmarks and a large-scale industrial ranking system. DADF consistently improves both pointwise accuracy and ranking quality across datasets and backbones. In the industrial setting, it achieves a 1.88 percentage-point WUAUC gain over the production baseline, reduces MAE by 12.57%, and yields a statistically significant 0.347% lift in average time spent per device in online A/B testing. These results demonstrate that DADF effectively mitigates local calibration bias and provides a practical plug-in solution for debiasing long-tailed continuous targets. The source code is available at https://github.com/liuzhao09/DADF.

关键词

引用

@article{arxiv.2605.17863,
  title  = {DADF: A Distribution-Aware Debiasing Framework for Watch-Time Regression in Recommender Systems},
  author = {Yiqing Yang and Xinlong Zhao and Zhao Liu and Xiao Lv and Ruiming Tang and Han Li and Kun Gai},
  journal= {arXiv preprint arXiv:2605.17863},
  year   = {2026}
}

备注

12 pages, 7 figures, 3 tables