Fact-Checking with Large Language Models via Probabilistic Certainty and Consistency

2026-01-05

arXiv.org

AI Summary

This work proposes an adaptive verification framework inspired by human fact-checking to mitigate hallucinations in large language models (LLMs) when generating factual content. Existing fact-checking approaches often rely on blind external retrieval, neglecting the model’s internal knowledge and introducing noise. In contrast, the proposed method dynamically quantifies the model’s factual confidence in a given claim by jointly modeling its probabilistic certainty and multi-path reasoning consistency. Based on this confidence estimate, the framework adaptively selects among three strategies: direct answering, targeted retrieval, or deep search—invoking external evidence only when necessary. Evaluated on three challenging benchmarks, the approach significantly outperforms strong baselines, demonstrating superior uncertainty quantification and cross-model generalization capabilities.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Large language models (LLMs) are increasingly used in applications requiring factual accuracy, yet their outputs often contain hallucinated responses. While fact-checking can mitigate these errors, existing methods typically retrieve external evidence indiscriminately, overlooking the model's internal knowledge and potentially introducing irrelevant noise. Moreover, current systems lack targeted mechanisms to resolve specific uncertainties in the model's reasoning. Inspired by how humans fact-check, we argue that LLMs should adaptively decide whether to rely on internal knowledge or initiate retrieval based on their confidence in a given claim. We introduce Probabilistic Certainty and Consistency (PCC), a framework that estimates factual confidence by jointly modeling an LLM's probabilistic certainty and reasoning consistency. These confidence signals enable an adaptive verification strategy: the model answers directly when confident, triggers targeted retrieval when uncertain or inconsistent, and escalates to deep search when ambiguity is high. Our confidence-guided routing mechanism ensures that retrieval is invoked only when necessary, improving both efficiency and reliability. Extensive experiments across three challenging benchmarks show that PCC achieves better uncertainty quantification than verbalized confidence and consistently outperforms strong LLM-based fact-checking baselines. Furthermore, we demonstrate that PCC generalizes well across various LLMs.

Problem

• fact-checking
• large language models
• hallucination
• uncertainty quantification
• retrieval

Innovation

• probabilistic certainty
• reasoning consistency
• adaptive retrieval
• confidence-guided routing
• fact-checking