![[2512.24787] HiGR: Efficient Generative Slate Recommendation via Hierarchical Planning and Multi-Objective Preference Alignment](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2512.24787] HiGR: Efficient Generative Slate Recommendation via Hierarchical Planning and Multi-Objective Preference Alignment
Summary
The paper presents HiGR, a novel framework for generative slate recommendation that enhances efficiency and user preference alignment through hierarchical planning.
Why It Matters
HiGR addresses critical limitations in current slate recommendation systems, such as inefficiencies in decoding and alignment with user preferences. Its deployment in a commercial setting demonstrates practical benefits, making it relevant for industries relying on recommendation systems.
Key Takeaways
- HiGR integrates hierarchical planning to improve slate recommendation efficiency.
- The framework achieves a 5x speedup in inference while enhancing recommendation quality.
- It utilizes multi-objective preference alignment to better match user needs.
- Experimental results show over 10% improvement in offline recommendation quality.
- HiGR has been successfully deployed in a commercial platform, yielding measurable user engagement increases.
Computer Science > Information Retrieval arXiv:2512.24787 (cs) [Submitted on 31 Dec 2025 (v1), last revised 23 Feb 2026 (this version, v2)] Title:HiGR: Efficient Generative Slate Recommendation via Hierarchical Planning and Multi-Objective Preference Alignment Authors:Yunsheng Pang, Zijian Liu, Yudong Li, Shaojie Zhu, Zijian Luo, Chenyun Yu, Sikai Wu, Shichen Shen, Cong Xu, Bin Wang, Kai Jiang, Hongyong Yu, Chengxiang Zhuo, Zang Li View a PDF of the paper titled HiGR: Efficient Generative Slate Recommendation via Hierarchical Planning and Multi-Objective Preference Alignment, by Yunsheng Pang and 13 other authors View PDF HTML (experimental) Abstract:Slate recommendation, which presents users with a ranked item list in a single display, is ubiquitous across mainstream online platforms. Recent advances in generative models have shown significant potential for this task via autoregressive modeling of discrete semantic ID sequences. However, existing methods suffer from three key limitations: entangled item tokenization, inefficient sequential decoding, and the absence of holistic slate planning. These issues often result in substantial inference overhead and inadequate alignment with diverse user preferences and practical business requirements, hindering the industrial deployment of generative slate recommendation systems. In this paper, we propose HiGR, an efficient generative slate recommendation framework that integrates hierarchical planning with listwise preference alig...
