# Relative Advantage Debiasing for Watch-Time Prediction in Short-Video Recommendation

arXiv: https://arxiv.org/abs/2508.11086 · HTML: https://mlsanity.com/paper/2508.11086/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 41 references that are in the map.
- First posted on arXiv 14 Aug 2025.

## Builds on (most cited first, 20 of 20)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [DCN V2: Improved Deep & Cross Network and Practical Lessons for Web-scale Learning to Rank Systems](https://mlsanity.com/paper/2008.13535/) (2020)
- [Group Robust Preference Optimization in Reward-free RLHF](https://mlsanity.com/paper/2405.20304/) (2024)
- [KuaiRand: An Unbiased Sequential Recommendation Dataset with Randomly Exposed Videos](https://mlsanity.com/paper/2208.08696/) (2022)
- [Optimized Cost per Click in Taobao Display Advertising](https://mlsanity.com/paper/1703.02091/) (2017)
- [Seq2Slate: Re-ranking and Slate Optimization with RNNs](https://mlsanity.com/paper/1810.02019/) (2018)
- [Deconfounding Duration Bias in Watch-time Prediction for Video Recommendation](https://mlsanity.com/paper/2206.06003/) (2022)
- [Beyond expectation: Deep joint mean and quantile regression for spatio-temporal problems](https://mlsanity.com/paper/1808.08798/) (2018)
- [Quantile regression with deep ReLU Networks: Estimators and minimax rates](https://mlsanity.com/paper/2010.08236/) (2020)
- [Generative Flow Network for Listwise Recommendation](https://mlsanity.com/paper/2306.02239/) (2023)
- [Concept-Aware Denoising Graph Neural Network for Micro-Video Recommendation](https://mlsanity.com/paper/2109.13527/) (2021)
- [Tree based Progressive Regression Model for Watch-Time Prediction in Short-video Recommendation](https://mlsanity.com/paper/2306.03392/) (2023)
- [CREAD: A Classification-Restoration Framework with Error Adaptive Discretization for Watch Time Prediction in Video Recommender Systems](https://mlsanity.com/paper/2401.07521/) (2024)
- [DVR: Micro-Video Recommendation Optimizing Watch-Time-Gain under Duration Bias](https://mlsanity.com/paper/2208.05190/) (2022)
- [Alleviating Video-Length Effect for Micro-video Recommendation](https://mlsanity.com/paper/2308.14276/) (2023)
- [Counteracting Duration Bias in Video Recommendation via Counterfactual Watch Time](https://mlsanity.com/paper/2406.07932/) (2024)
- [Leveraging Watch-time Feedback for Short-Video Recommendations: A Causal Labeling Framework](https://mlsanity.com/paper/2306.17426/) (2023)
- [Conditional Quantile Estimation for Uncertain Watch Time in Short-Video Recommendation](https://mlsanity.com/paper/2407.12223/) (2024)
- [Nonparametric Quantile Regression: Non-Crossing Constraints and Conformal Prediction](https://mlsanity.com/paper/2210.10161/) (2022)
- [AlignPxtr: Aligning Predicted Behavior Distributions for Bias-Free Video Recommendations](https://mlsanity.com/paper/2503.06920/) (2025)

## Built on by (most cited first, 4 of 4)

- [Hierarchical Exponential-Gaussian Mixtures for Watch-Time Distribution Prediction](https://mlsanity.com/paper/2608.23356/) (2026)
- [PEARL: Unbiased Percentile Estimation via Contrastive Learning for Industrial-Scale Livestream Recommendation](https://mlsanity.com/paper/2605.21752/) (2026)
- [PIT-SUN: A Deployable Empirical Marginal Transform Framework with Expectation-Consistent Recovery for Regression in Recommender Systems](https://mlsanity.com/paper/2607.08202/) (2026)
- [Mend the Measurement Gap: Latent User Preference Modeling for Short-Form Video Recommendation](https://mlsanity.com/paper/2609.32839/) (2026)

## Abstract (arXiv)

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.