Towards Unification of Hallucination Detection and Fact Verification for Large Language Models

2025-12-02

AI Summary

Hallucination in large language models (LLMs) severely undermines their reliability and real-world deployability. Current research is fragmented into two incompatible paradigms: model-centric hallucination detection (HD) and text-centric fact verification (FV), differing fundamentally in assumptions, data construction, and evaluation protocols. To bridge this gap, we propose UniFact—the first unified evaluation framework enabling instance-level, directly comparable assessment of both HD and FV. Leveraging dynamically generated–annotated paired data, we systematically uncover their complementary strengths and demonstrate that hybrid HD+FV methods substantially outperform either paradigm alone, achieving new state-of-the-art performance. We open-source all code, datasets, and baseline systems to foster convergence in hallucination research, establishing unified evaluation and collaborative modeling as the new standard.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models (LLMs) frequently exhibit hallucinations, generating content that appears fluent and coherent but is factually incorrect. Such errors undermine trust and hinder their adoption in real-world applications. To address this challenge, two distinct research paradigms have emerged: model-centric Hallucination Detection (HD) and text-centric Fact Verification (FV). Despite sharing the same goal, these paradigms have evolved in isolation, using distinct assumptions, datasets, and evaluation protocols. This separation has created a research schism that hinders their collective progress. In this work, we take a decisive step toward bridging this divide. We introduce UniFact, a unified evaluation framework that enables direct, instance-level comparison between FV and HD by dynamically generating model outputs and corresponding factuality labels. Through large-scale experiments across multiple LLM families and detection methods, we reveal three key findings: (1) No paradigm is universally superior; (2) HD and FV capture complementary facets of factual errors; and (3) hybrid approaches that integrate both methods consistently achieve state-of-the-art performance. Beyond benchmarking, we provide the first in-depth analysis of why FV and HD diverged, as well as empirical evidence supporting the need for their unification. The comprehensive experimental results call for a new, integrated research agenda toward unifying Hallucination Detection and Fact Verification in LLMs. We have open-sourced all the code, data, and baseline implementation at: https://github.com/oneal2000/UniFact/

Problem

• Unifies hallucination detection and fact verification for LLMs
• Enables direct comparison between two isolated research paradigms
• Demonstrates hybrid methods achieve best performance

Innovation

• Unified evaluation framework for hallucination detection and fact verification
• Dynamic generation of model outputs with factuality labels
• Hybrid approaches combining both methods achieve best performance