Chapter 10 · 2025
Zero-knowledge LLM hallucination detection and mitigation through fine-grained cross-model consistency
Zexin Li, Zuchao Li, Xinyu Li
Abstract
Large Language Models (LLMs) are increasingly deployed in real-world applications, yet their propensity for generating factually incorrect or inconsistent information, known as hallucination, remains a significant concern. Current methods for hallucination detection often rely on external knowledge bases or require extensive human annotation, which can be costly and time-consuming. This paper proposes a novel zero-knowledge approach for LLM hallucination detection and mitigation based on fine-grained cross-model consistency. Our method leverages the inherent variability in responses from different LLMs to identify inconsistencies at a granular level, without requiring access to ground truth or external factual sources.
Topics
LLM HallucinationZero-knowledge DetectionCross-model ConsistencyHallucination MitigationFine-grained Analysis
Relevance Scores
Long-Horizon Score85
Enterprise Score80
Completeness75
Paper Info
Year2025
Venue
Type
ChapterCh. 10
Authors3
Zone III Analysis
Frameworks
Related Papers
ReAct: Synergizing Reasoning and Acting in Language Mod…
2023 · Ch.1
Toolformer: Language Models Can Teach Themselves to Use…
2023 · Ch.1
MemGPT: Towards LLMs as Operating Systems
2023 · Ch.2
Durable Execution: Temporal and Cadence-Aware Workflow …
2023 · Ch.2
View all Chapter 10 papers →