Confidence in the Reasoning of Large Language Models

TL;DR

This paper investigates how large language models assess their own confidence in answers, revealing that they tend to overestimate confidence and lack an internally coherent sense of certainty, despite performing better than random guessing.

Contribution

It introduces a novel evaluation of LLM confidence through qualitative persistence and self-reported scores, highlighting limitations in their self-assessment capabilities.

Findings

LLMs perform better than random guessing on reasoning tasks.

There is a positive correlation between qualitative confidence and accuracy.

LLMs tend to overstate their confidence scores.

Abstract

There is a growing literature on reasoning by large language models (LLMs), but the discussion on the uncertainty in their responses is still lacking. Our aim is to assess the extent of confidence that LLMs have in their answers and how it correlates with accuracy. Confidence is measured (i) qualitatively in terms of persistence in keeping their answer when prompted to reconsider, and (ii) quantitatively in terms of self-reported confidence score. We investigate the performance of three LLMs -- GPT4o, GPT4-turbo and Mistral -- on two benchmark sets of questions on causal judgement and formal fallacies and a set of probability and statistical puzzles and paradoxes. Although the LLMs show significantly better performance than random guessing, there is a wide variability in their tendency to change their initial answers. There is a positive correlation between qualitative confidence and accuracy, but the overall accuracy for the second answer is often worse than for the first answer. There is a strong tendency to overstate the self-reported confidence score. Confidence is only partially explained by the underlying token-level probability. The material effects of prompting on qualitative confidence and the strong tendency for overconfidence indicate that current LLMs do not have any internally coherent sense of confidence.