Papers
arxiv:2512.23511

Beyond Correctness: Exposing LLM-generated Logical Flaws in Reasoning via Multi-step Automated Theorem Proving

Published on Dec 29, 2025
Authors:
,
,
,
,
,
,

Abstract

MATP is an evaluation framework that uses multi-step automatic theorem proving to verify large language model reasoning by translating natural language to first-order logic and detecting logical errors.

Large Language Models (LLMs) have demonstrated impressive reasoning capabilities, leading to their adoption in high-stakes domains such as healthcare, law, and scientific research. However, their reasoning often contains subtle logical errors masked by fluent language, posing significant risks for critical applications. While existing approaches like fact-checking, self-consistency methods, and rule-based validation provide partial solutions, they fail to detect complex logical flaws in multi-step reasoning. To overcome these challenges, we present MATP, an evaluation framework for systematically verifying LLM reasoning via Multi-step Automatic Theorem Proving. MATP translates natural language reasoning into First-Order Logic (FOL) and applies automated theorem provers to assess step-by-step logical validity. This approach identifies hidden logical errors and provides fine-grained classifications of reasoning correctness. Evaluations on a benchmark comprising 10,830 reasoning instances generated by 10 LLMs across tasks from PrOntoQA-OOD, ProofWriter, and FOLIO show that MATP surpasses prompting-based baselines by over 42 percentage points in reasoning step verification. It further reveals model-level disparities, with reasoning models generating more logically coherent outputs than general models. These results demonstrate MATP's potential to enhance the trustworthiness of LLM-generated reasoning.

Community

Sign up or log in to comment

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2512.23511 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2512.23511 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2512.23511 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.