Logical Consistency as a Bridge: Improving LLM Hallucination Detection via Label Constraint Modeling between Responses and Self-Judgments

2026-05-05

AI Summary

This work addresses the prevalent issue of factual hallucinations in large language models (LLMs), noting that existing detection approaches rely solely on either neural uncertainty or symbolic self-judgment while neglecting their interplay. To bridge this gap, the authors propose the Logic-aware Alignment via Bridge (LaaB) framework, which establishes a logical consistency link between model responses and self-judgments by mapping symbolic labels back into the feature space, thereby unifying neural and symbolic perspectives. The core innovation lies in a novel “meta-judgment” mechanism that leverages semantic constraints between responses and meta-judgments to align and mutually reinforce neural representations and symbolic reasoning. Integrating uncertainty quantification, self-judgment prompt engineering, and multi-view mutual learning, LaaB consistently outperforms eight baselines across four datasets and four LLMs, demonstrating strong effectiveness and generalization capability.

Abstract

Large Language Models (LLMs) are prone to factual hallucinations, risking their reliability in real-world applications. Existing hallucination detectors mainly extract micro-level intrinsic patterns for uncertainty quantification or elicit macro-level self-judgments through verbalized prompts. However, these methods address only a single facet of the hallucination, focusing either on implicit neural uncertainty or explicit symbolic reasoning, thereby treating these inherently coupled behaviors in isolation and failing to exploit their interdependence for a holistic view. In this paper, we propose LaaB (Logical Consistency-as-a-Bridge), a framework that bridges neural features and symbolic judgments for hallucination detection. LaaB introduces a "meta-judgment" process to map symbolic labels back into the feature space. By leveraging the inherent logical bridge where response and meta-judgment labels are either the same or opposite based on the self-judgment's semantics, LaaB aligns and integrates dual-view signals via mutual learning and enhances the hallucination detection. Extensive experiments on 4 public datasets, across 4 LLMs, against 8 baselines demonstrate the superiority of LaaB.

Problem

• hallucination detection
• logical consistency
• large language models
• self-judgment
• neural-symbolic integration

Innovation

• logical consistency
• hallucination detection
• self-judgment
• meta-judgment
• mutual learning