Sound and Complete Neurosymbolic Reasoning with LLM-Grounded Interpretations

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Allen, Bradley P., Chhikara, Prateek, Ferguson, Thomas Macaulay, Ilievski, Filip, Groth, Paul
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918110257217536
author Allen, Bradley P.
Chhikara, Prateek
Ferguson, Thomas Macaulay
Ilievski, Filip
Groth, Paul
author_facet Allen, Bradley P.
Chhikara, Prateek
Ferguson, Thomas Macaulay
Ilievski, Filip
Groth, Paul
contents Large language models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation, but they exhibit problems with logical consistency in the output they generate. How can we harness LLMs' broad-coverage parametric knowledge in formal reasoning despite their inconsistency? We present a method for directly integrating an LLM into the interpretation function of the formal semantics for a paraconsistent logic. We provide experimental evidence for the feasibility of the method by evaluating the function using datasets created from several short-form factuality benchmarks. Unlike prior work, our method offers a theoretical framework for neurosymbolic reasoning that leverages an LLM's knowledge while preserving the underlying logic's soundness and completeness properties.
format Preprint
id arxiv_https___arxiv_org_abs_2507_09751
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Sound and Complete Neurosymbolic Reasoning with LLM-Grounded Interpretations
Allen, Bradley P.
Chhikara, Prateek
Ferguson, Thomas Macaulay
Ilievski, Filip
Groth, Paul
Artificial Intelligence
Computation and Language
Logic in Computer Science
Large language models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation, but they exhibit problems with logical consistency in the output they generate. How can we harness LLMs' broad-coverage parametric knowledge in formal reasoning despite their inconsistency? We present a method for directly integrating an LLM into the interpretation function of the formal semantics for a paraconsistent logic. We provide experimental evidence for the feasibility of the method by evaluating the function using datasets created from several short-form factuality benchmarks. Unlike prior work, our method offers a theoretical framework for neurosymbolic reasoning that leverages an LLM's knowledge while preserving the underlying logic's soundness and completeness properties.
title Sound and Complete Neurosymbolic Reasoning with LLM-Grounded Interpretations
topic Artificial Intelligence
Computation and Language
Logic in Computer Science
url https://arxiv.org/abs/2507.09751