Search papers, labs, and topics across Lattice.
This paper introduces the Hierarchical Exponential-Gaussian Mixture (HEGM) model for predicting watch-time distributions in short-video recommendations, addressing the limitations of the Exponential-Gaussian Mixture Network (EGMN) which suffers from variance collapse and component redundancy. By employing a hierarchical skip-watch decomposition and KL-based variance regularization, HEGM enhances both the stability and interpretability of the mixture model while achieving superior ranking accuracy and threshold-event prediction. A real-world A/B test over 1.5 months demonstrates statistically significant engagement improvements, validating the practical impact of the proposed model.
HEGM not only stabilizes watch-time predictions but also boosts user engagement in real-world applications, outperforming its predecessor in both accuracy and interpretability.
Accurate watch-time (WT) prediction is an important requirement for short-video recommendations. Yet WT distributions are near-zero-inflated, long-tailed and multimodal. The recent Exponential-Gaussian Mixture Network (EGMN) models the full conditional WT distribution rather than a single point estimate and achieves state-of-the-art performance. Our large-scale reproduction study reveals that EGMN is vulnerable to variance collapse, component redundancy, and inactive components. We propose a Hierarchical Exponential-Gaussian Mixture (HEGM) model that addresses these failure modes through a hierarchical skip-watch decomposition, KL-based variance regularization, structured initialization, removing the forced Gaussian shift and the entropy regularizer. Across public and large-scale industrial datasets, HEGM improves ranking accuracy and threshold-event prediction, while maintaining competitive point-estimation accuracy and substantially improving mixture stability and interpretability. A 1.5-month production A/B test confirms statistically significant engagement lifts. Our code and models are publicly released at https://github.com/rw404/HEGM.
