Towards Logically Sound Natural Language Reasoning with Logic-Enhanced Language Model Agents

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Mensfelt, Agnieszka, Stathis, Kostas, Trencsenyi, Vince
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866918037858287616
author Mensfelt, Agnieszka
Stathis, Kostas
Trencsenyi, Vince
author_facet Mensfelt, Agnieszka
Stathis, Kostas
Trencsenyi, Vince
contents Large language models (LLMs) are increasingly explored as general-purpose reasoners, particularly in agentic contexts. However, their outputs remain prone to mathematical and logical errors. This is especially challenging in open-ended tasks, where unstructured outputs lack explicit ground truth and may contain subtle inconsistencies. To address this issue, we propose Logic-Enhanced Language Model Agents (LELMA), a framework that integrates LLMs with formal logic to enable validation and refinement of natural language reasoning. LELMA comprises three components: an LLM-Reasoner, an LLM-Translator, and a Solver, and employs autoformalization to translate reasoning into logic representations, which are then used to assess logical validity. Using game-theoretic scenarios such as the Prisoner's Dilemma as testbeds, we highlight the limitations of both less capable (Gemini 1.0 Pro) and advanced (GPT-4o) models in generating logically sound reasoning. LELMA achieves high accuracy in error detection and improves reasoning correctness via self-refinement, particularly in GPT-4o. The study also highlights challenges in autoformalization accuracy and in evaluation of inherently ambiguous open-ended reasoning tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2408_16081
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Towards Logically Sound Natural Language Reasoning with Logic-Enhanced Language Model Agents
Mensfelt, Agnieszka
Stathis, Kostas
Trencsenyi, Vince
Artificial Intelligence
Computation and Language
Computer Science and Game Theory
Logic in Computer Science
Large language models (LLMs) are increasingly explored as general-purpose reasoners, particularly in agentic contexts. However, their outputs remain prone to mathematical and logical errors. This is especially challenging in open-ended tasks, where unstructured outputs lack explicit ground truth and may contain subtle inconsistencies. To address this issue, we propose Logic-Enhanced Language Model Agents (LELMA), a framework that integrates LLMs with formal logic to enable validation and refinement of natural language reasoning. LELMA comprises three components: an LLM-Reasoner, an LLM-Translator, and a Solver, and employs autoformalization to translate reasoning into logic representations, which are then used to assess logical validity. Using game-theoretic scenarios such as the Prisoner's Dilemma as testbeds, we highlight the limitations of both less capable (Gemini 1.0 Pro) and advanced (GPT-4o) models in generating logically sound reasoning. LELMA achieves high accuracy in error detection and improves reasoning correctness via self-refinement, particularly in GPT-4o. The study also highlights challenges in autoformalization accuracy and in evaluation of inherently ambiguous open-ended reasoning tasks.
title Towards Logically Sound Natural Language Reasoning with Logic-Enhanced Language Model Agents
topic Artificial Intelligence
Computation and Language
Computer Science and Game Theory
Logic in Computer Science
url https://arxiv.org/abs/2408.16081