Exploring the Meta-level Reasoning of Large Language Models via a Tool-based Multi-hop Tabular Question Answering Task

TL;DR

This paper introduces a structured multi-hop question answering task to evaluate the meta-level reasoning abilities of large language models, focusing on tool selection and intermediate reasoning steps, revealing strengths and limitations in their reasoning and numeracy skills.

Contribution

It presents a novel task for assessing meta-level reasoning in LLMs, emphasizing tool use and intermediate step analysis, advancing understanding beyond final answer accuracy.

Findings

LLMs show good meta-level reasoning on the task

N-shot prompting has little impact on accuracy

LLMs exhibit poor numeracy skills

Abstract

Recent advancements in Large Language Models (LLMs) are increasingly focused on "reasoning" ability, a concept with many overlapping definitions in the LLM discourse. We take a more structured approach, distinguishing meta-level reasoning (denoting the process of reasoning about intermediate steps required to solve a task) from object-level reasoning (which concerns the low-level execution of the aforementioned steps.) We design a novel question answering task, which is based around the values of geopolitical indicators for various countries over various years. Questions require breaking down into intermediate steps, retrieval of data, and mathematical operations over that data. The meta-level reasoning ability of LLMs is analysed by examining the selection of appropriate tools for answering questions. To bring greater depth to the analysis of LLMs beyond final answer accuracy, our task contains 'essential actions' against which we can compare the tool call output of LLMs to infer the strength of reasoning ability. We find that LLMs demonstrate good meta-level reasoning on our task, yet are flawed in some aspects of task understanding. We find that n-shot prompting has little effect on accuracy; error messages encountered do not often deteriorate performance; and provide additional evidence for the poor numeracy of LLMs. Finally, we discuss the generalisation and limitation of our findings to other task domains.