Prospects for inconsistency detection using large language models and sheaves

TL;DR

This paper explores how large language models can assess the logical consistency of claims and proposes a sheaf-theoretic mathematical framework to evaluate consistency across complex hypertexts, aiming to improve societal information integrity.

Contribution

It introduces a novel combination of large language models with sheaf theory to evaluate and enhance consistency in complex textual hyperstructures.

Findings

Language models can reasonably rate claim consistency.

Sheaf theory provides a mathematical framework for global consistency evaluation.

Potential applications in law, social media, and governance.

Abstract

We demonstrate that large language models can produce reasonable numerical ratings of the logical consistency of claims. We also outline a mathematical approach based on sheaf theory for lifting such ratings to hypertexts such as laws, jurisprudence, and social media and evaluating their consistency globally. This approach is a promising avenue to increasing consistency in and of government, as well as to combating mis- and disinformation and related ills.