English

On Evaluating the Integration of Reasoning and Action in LLM Agents with Database Question Answering

Computation and Language 2023-11-17 v1

Abstract

This study introduces a new long-form database question answering dataset designed to evaluate how Large Language Models (LLMs) interact with a SQL interpreter. The task necessitates LLMs to strategically generate multiple SQL queries to retrieve sufficient data from a database, to reason with the acquired context, and to synthesize them into a comprehensive analytical narrative. Our findings highlight that this task poses great challenges even for the state-of-the-art GPT-4 model. We propose and evaluate two interaction strategies, and provide a fine-grained analysis of the individual stages within the interaction. A key discovery is the identification of two primary bottlenecks hindering effective interaction: the capacity for planning and the ability to generate multiple SQL queries. To address the challenge of accurately assessing answer quality, we introduce a multi-agent evaluation framework that simulates the academic peer-review process, enhancing the precision and reliability of our evaluations. This framework allows for a more nuanced understanding of the strengths and limitations of current LLMs in complex retrieval and reasoning tasks.

Keywords

Cite

@article{arxiv.2311.09721,
  title  = {On Evaluating the Integration of Reasoning and Action in LLM Agents with Database Question Answering},
  author = {Linyong Nan and Ellen Zhang and Weijin Zou and Yilun Zhao and Wenfei Zhou and Arman Cohan},
  journal= {arXiv preprint arXiv:2311.09721},
  year   = {2023}
}