Personalized recommendation method improves user engagement and diversity
SPAR: Stepwise Personalized Autoregressive Recommendation with Prefix-Balanced Semantic IDs
Information Retrieval
Summary
Recommender systems suggest items like videos or products based on what a user likes, but they can make mistakes if early guesses go wrong or if popular items get too much focus. The authors propose SPAR, a new approach that improves step-by-step decision making by looking at past user behavior and balancing the choices more fairly. This method was tested on a large video platform and showed better accuracy, more balanced suggestions, and higher user watch time compared to existing systems.
What this means in practice
- •For video streaming platform engineers: Improve video recommendations by better balancing user history and current preferences to increase watch time and engagement.
- •For e-commerce recommendation teams: Develop product suggestion systems that reduce bias toward popular items by balancing choices at each decision step using user activity data.
Authors
Chunrui Zhang, Gaoyun Lin, Chang Zhou, Zheng Hu, Qiran Song, Shimin Cai, Tao Zhou
Abstract
Semantic ID (SID)-based generative recommendation retrieves items through successive token-level decisions, making personalization at each decoding step essential. However, existing methods insufficiently account for prefix-dependent user evidence and imbalanced historical interaction loads among candidate branches. This can cause irreversible branch-selection errors and favor popular paths. We propose SPAR, a generative recommendation framework that jointly improves stepwise personalization and SID construction. Its PATH Decoder retrieves prefix-relevant evidence from compressed long-term user memories and combines it with recent behavior, user profiles, and event context to guide each token decision. Its RQ-PIB module balances historical interaction loads among sibling branches under each parent prefix while preserving semantic structure. Experiments on large-scale Youku production data show an average relative HitRate improvement of 2.66% over the strongest baseline. SPAR also increases target-prefix retention by 5.79 percentage points and reduces the traffic-weighted largest child-branch share by 32.84% across SID levels. In a 7-day online A/B test, SPAR serves 10% of Youku's production traffic and improves average watch time per user by 3.95% and play conversion rate by 3.38%. These results demonstrate more reliable hierarchical decoding, balanced SID branches, and improved user engagement in industrial recommendation.