Efficient Optimization of Hierarchical Identifiers for Generative Recommendation

2025-12-20

AI Summary

To address the high computational cost and poor scalability of tree construction in the SEATER generative recommendation model under large-scale item scenarios, this paper proposes an efficient hierarchical identifier construction method. Our approach introduces greedy and hybrid hierarchical tree-building algorithms, integrates balanced tree structural modeling, a contrastive learning objective function, and multi-granularity grouping strategies—collectively accelerating tree construction significantly. Experiments on large-scale music datasets (e.g., Yambda) demonstrate that our method reduces tree construction time to only 2%–8% of the original SEATER’s runtime, while maintaining or even improving retrieval quality. All code and datasets are publicly released to ensure full reproducibility.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Feature Construction/Reformulation
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

SEATER is a generative retrieval model that improves recommendation inference efficiency and retrieval quality by utilizing balanced tree-structured item identifiers and contrastive training objectives. We reproduce and validate SEATER's reported improvements in retrieval quality over strong baselines across all datasets from the original work, and extend the evaluation to Yambda, a large-scale music recommendation dataset. Our experiments verify SEATER's strong performance, but show that its tree construction step during training becomes a major bottleneck as the number of items grows. To address this, we implement and evaluate two alternative construction algorithms: a greedy method optimized for minimal build time, and a hybrid method that combines greedy clustering at high levels with more precise grouping at lower levels. The greedy method reduces tree construction time to less than 2% of the original with only a minor drop in quality on the dataset with the largest item collection. The hybrid method achieves retrieval quality on par with the original, and even improves on the largest dataset, while cutting construction time to just 5-8%. All data and code are publicly available for full reproducibility at https://github.com/joshrosie/re-seater.

Problem

• Optimizes hierarchical identifiers for generative recommendation efficiency
• Addresses tree construction bottleneck in large-scale item datasets
• Evaluates alternative algorithms to reduce training time while maintaining quality

Innovation

• Balanced tree-structured identifiers improve retrieval efficiency
• Contrastive training objectives enhance recommendation quality
• Hybrid tree construction algorithm reduces training time significantly