arXiv cs.AI· Fuheng Zhao, Jiayue Chen, Yiming Pan, Tahseen Rabbani, Sohaib, Divyakant Agrawal, Amr El Abbadi, Paritosh Aggarwal, Anupam Datta, Dimitris Tsirogiannis·· 6 天前AI 评分21
面向 LLM ORDER BY 高效排序的访问路径研究
Access Paths for Efficient Ordering with Large Language Models
AI 导读
研究提出 LLM ORDER BY 语义算子，并系统比较其物理实现，包括改进现有语义排序算法与新增语义感知外部归并排序。评估显示没有单一实现在所有数据集上最优，且基于比较的算法存在排序成本与排序质量的缩放关系。为此设计的预算感知优化器结合启发式规则、LLM-as-Judge 与共识聚合动态选择访问路径，在所有基准上排名准确率持平或优于最佳静态方法。
正文 · 原文
Abstract:In this work, we present the \texttt{LLM ORDER BY} semantic operator as a logical abstraction and conduct a systematic study of its physical implementations. First, we propose several improvements to existing semantic sorting algorithms and introduce a semantic-aware external merge sort algorithm. Our extensive evaluation reveals that no single implementation offers universal optimality on all datasets. From our evaluations, we observe a general scaling relationship between sorting cost and the ordering quality for comparison-based algorithms. Building on these insights, we design a budget-aware optimizer that utilizes heuristic rules, LLM-as-Judge evaluation, and consensus aggregation to dynamically select the near-optimal access path for LLM ORDER BY. In our extensive evaluations, our optimizer consistently achieves ranking accuracy on par with or superior to the best static methods across all benchmarks. We believe that this work provides foundational insights into the principled optimization of semantic operators essential for building robust, large-scale LLM-powered analytic systems.
| Subjects: | Databases (cs.DB); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR) |
| Cite as: | arXiv:2509.00303 [cs.DB] |
| (or arXiv:2509.00303v5 [cs.DB] for this version) | |
| https://doi.org/10.48550/arXiv.2509.00303 arXiv-issued DOI via DataCite |
Submission history
From: Fuheng Zhao [view email]
[v1]
Sat, 30 Aug 2025 01:44:36 UTC (218 KB)
[v2]
Wed, 3 Dec 2025 06:35:17 UTC (831 KB)
[v3]
Tue, 19 May 2026 23:05:35 UTC (1,189 KB)
[v4]
Mon, 14 Sep 2026 05:13:18 UTC (1,868 KB)
[v5]
Fri, 9 Oct 2026 15:39:40 UTC (1,882 KB)
来源：arXiv cs.AI · arxiv.org