Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

QUITE: A Query Rewrite System Beyond Rules with LLM Agents

Created by
  • Haebom

Author

Yuyang Song, Hanxu Yan, Jiale Lao, Yibo Wang, Yufei Li, Yuanchun Zhou, Jianguo Wang, Mingjie Tang

Outline

In this paper, we propose QUITE, a novel approach that leverages a large-scale language model (LLM) to overcome the limitations of existing rule-based SQL query rewriting methods. Existing methods rely on predefined rules, which limits the diversity of query patterns and rewriting strategies and can cause performance degradation. QUITE is a learning-free feedback-aware system based on LLM agents. It solves the hallucination problem of LLM, supports a wider range of query patterns and rewriting strategies, and improves performance through a multi-agent framework using a finite state machine (FSM), a rewriting middleware, and a hint injection technique. Experimental results show that QUITE reduces query execution time by up to 35.8% compared to state-of-the-art methods, and rewrites 24.1% more queries than existing methods.

Takeaways, Limitations

•
Takeaways:
◦
We present a novel SQL query rewriting method that overcomes the limitations of existing rule-based methods by leveraging LLM.
◦
Greatly improves the efficiency of query optimization by supporting various query patterns and rewriting strategies.
◦
Experimentally demonstrated performance improvements with up to 35.8% reduction in query execution time and 24.1% increase in the number of rewritten queries.
◦
Effectively solving LLM's hallucination problem using FSM-based multi-agent framework, rewriting middleware, and hint injection technique.
•
Limitations:
◦
Since it depends on the performance of LLM, the performance of QUITE may also be affected as the performance of LLM improves.
◦
Optimization may be required for specific database systems or query types.
◦
Additional adaptability evaluations for new query patterns or rewriting strategies may be required.
◦
The complexity of the design of FSMs and the interactions between agents can make system maintenance and scalability difficult.
👍
Made with Slashpage