POEM: Partial-Order Enhanced Real-Time Sequential Modeling for Recommendation

2026-06-29

AI Summary

This work addresses the limitations of existing real-time recommender systems in capturing dynamic user interest drift and contextual shifts, as conventional sequential models rely solely on static click behaviors and overlook structured signals inherent in multi-stage ranking pipelines. To overcome this, the authors propose POEM, a novel framework that constructs dynamic partial-order sequences from real-time scores generated by upstream multi-task ranking modules (e.g., CTR and watch time predictors) to enable fine-grained interest modeling. Key innovations include a partial-order-guided sequence construction mechanism based on ranking scores, a quintuple representation module integrating heterogeneous signals, and a hierarchical learning strategy that jointly leverages high-ranked items, positive feedback, and graph-mined hard negative samples. Deployed at scale on Kuaishou, POEM significantly enhances user engagement, increasing per-user watch time by 0.249% on KS Single Page and 0.213% on Light Page.

Abstract

Real-time recommendation systems suffer from the dynamic drift of user interests and varying contextual conditions. Conventional sequential recommendation models only exploit static historical click sequences, which fail to capture instant preference changes and overlook structured signals hidden within the multi-stage ranking pipeline of industrial recommendation systems. To tackle these limitations, we propose POEM (Partial-Order Enhanced Modeling), a new real-time sequential modeling framework built upon intrinsic partial-order relations from the recommendation cascade. POEM takes real-time multi-task ranking scores (including predicted CTR and predicted watch duration) generated by upstream ranking modules as supervision to construct dynamic partial-order sequences, supporting fine-grained real-time interest modeling and consistent optimization between system ranking targets and user behavioral patterns. We summarize our core contributions as three aspects: (1) a partial-order guided sequence construction paradigm, which enriches vanilla chronological sequences via dynamic grouping and sampling conditioned on real-time ranking scores to reassess user interests per request; (2) a multi-objective score fusion module that unifies heterogeneous ranking signals into a compact quintuple representation with normalized rank-aware weighting; (3) a hierarchical sample learning strategy, which adopts system-favored high-ranked items and user positive feedback (e.g., long-duration watched videos) as positive instances, paired with graph-mined hard negatives and a margin-based pairwise loss for robust training. Fully deployed on Kuaishou online traffic, POEM achieves significant online gains: average per-user watch time lifts by 0.249% on the KS Single Page and 0.213% on the KS Lite Page.

Problem

• real-time recommendation
• user interest drift
• sequential modeling
• partial-order relations
• multi-stage ranking

Innovation

• partial-order modeling
• real-time sequential recommendation
• multi-task ranking fusion
• hierarchical sample learning
• dynamic interest modeling