Exploring the potential for large language models to demonstrate rational probabilistic beliefs
File(s)
Published version
Author(s)
Freedman, Gabriel
Toni, Francesca
Type
Conference Paper
Abstract
Advances in the general capabilities of large language models (LLMs) have led to their use for information retrieval, and as components in automated decision systems. A faithful representation of probabilistic reasoning in these models may be essential to ensure trustworthy, explainable and effective performance in these tasks. Despite previous work suggesting that LLMs can perform complex reasoning and well-calibrated uncertainty quantification, we find that current versions of this class of model lack the ability to provide rational and coherent representations of probabilistic beliefs. To demonstrate this, we introduce a novel dataset of claims with indeterminate truth values and apply a number of well-established techniques for uncertainty quantification to measure the ability of LLM's to adhere to fundamental properties of probabilistic reasoning.
Date Issued
2025-05-14
Date Acceptance
2025-04-07
Citation
The International FLAIRS Conference Proceedings, 2025, 38 (1)
ISSN
2334-0762
Publisher
LibraryPress@UF
Journal / Book Title
The International FLAIRS Conference Proceedings
Volume
38
Issue
1
Copyright Statement
© 2025 Gabriel Freedman, Francesca Toni. This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License (https://creativecommons.org/licenses/by-nc/4.0/)
License URL
Source
38th International FLAIRS Conference
Publication Status
Published
Start Date
2025-05-20
Finish Date
2025-05-23
Coverage Spatial
Florida, USA