Explicit Uncertainty Modeling for Video Watch Time Prediction

2025-04-10

AI Summary

In video recommendation, predicting user watch time faces a fundamental trade-off between accuracy and out-of-distribution (OOD) generalization due to behavioral stochasticity; meanwhile, prediction uncertainty itself encodes valuable behavioral signals. To address this, we propose the first paradigm that explicitly models uncertainty as a learnable signal—integrating adversarial training, watch-time bias correction, and distribution-aware prediction. Our method resolves the accuracy–generalization dilemma without restricting backbone architectures. Evaluated on two public benchmarks, it significantly improves both robustness and accuracy of watch-time predictions. Deployed in production, online A/B testing demonstrates a 0.31% increase in total user watch time, validating the practical impact of explicit uncertainty modeling on recommendation performance.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Game Theory and Economic Paradigms: Adversarial Learning
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

In video recommendation, a critical component that determines the system's recommendation accuracy is the watch-time prediction module, since how long a user watches a video directly reflects personalized preferences. One of the key challenges of this problem is the user's stochastic watch-time behavior. To improve the prediction accuracy for such an uncertain behavior, existing approaches show that one can either reduce the noise through duration bias modeling or formulate a distribution modeling task to capture the uncertainty. However, the uncontrolled uncertainty is not always equally distributed across users and videos, inducing a balancing paradox between the model accuracy and the ability to capture out-of-distribution samples. In practice, we find that the uncertainty of the watch-time prediction model also provides key information about user behavior, which, in turn, could benefit the prediction task itself. Following this notion, we derive an explicit uncertainty modeling strategy for the prediction model and propose an adversarial optimization framework that can better exploit the user watch-time behavior. This framework has been deployed online on an industrial video sharing platform that serves hundreds of millions of daily active users, which obtains a significant increase in users' video watch time by 0.31% through the online A/B test. Furthermore, extended offline experiments on two public datasets verify the effectiveness of the proposed framework across various watch-time prediction backbones.

Problem

• Modeling user stochastic watch-time behavior uncertainty
• Balancing model accuracy and out-of-distribution sample capture
• Exploiting uncertainty to improve watch-time prediction accuracy

Innovation

• Explicit uncertainty modeling for watch-time prediction
• Adversarial optimization framework for behavior exploitation
• Deployed on industrial platform with significant user impact