Deep Learning Approaches for Hallucination Detection in Large Language Models: A Comprehensive Survey
IOSR Journal of Computer Engineering · 2026-03 · 28 (3) · 18-30
DOI: 10.9790/0661-2803041830 ↗Kurz gefasst
Fachartikel in der Zeitschrift IOSR Journal of Computer Engineering, 2026-03. Verfasst von Redwana Farbin und 3 weiteren Autor:innen.
Abstract (Originalsprache)
The rapid advancement of Large Language Models (LLMs) has facilitated remarkable progress in natural language generation. Nevertheless, a major challenge remains, and this is the problem of hallucinations, where the models generate outputs that are linguistically valid but factually incorrect, logically unjustified, and even completely fabricated. The consequences of undetected hallucinations are severe, spanning patient safety risks in healthcare, fabricated legal precedents in judicial contexts, and distorted scholarly discourse in scientific research. Although significant research has focused on understanding and mitigating hallucinations in large language models, comparatively fewer studies have systematically examined deep learning-based detection methods and their associated benchmark datasets. This survey addresses that gap by presenting a structured taxonomy of detection methods organized into five paradigms: confidence and uncertainty-based detection, consistency-based detection, knowledge-grounded and graph-based detection, auxiliary model-based detection, and internal representation-based detection. For each paradigm, we analyze representative methods, datasets and outcomes. The insights synthesized in this survey provide guidance for designing reliable hallucination-aware LLM systems and support the development of adaptive detection–mitigation frameworks for trustworthy largescale deployment.
Zitieren
APA
Farbin, R., Ismail, S., Paul, S. K., & Hamid, M. E. (2026). Deep Learning Approaches for Hallucination Detection in Large Language Models: A Comprehensive Survey. IOSR Journal of Computer Engineering, 28(3), 18-30. https://doi.org/10.9790/0661-2803041830
Vancouver
Farbin R, Ismail S, Paul SK, Hamid ME. Deep Learning Approaches for Hallucination Detection in Large Language Models: A Comprehensive Survey. IOSR Journal of Computer Engineering. 2026;28(3):18-30. doi:10.9790/0661-2803041830
BibTeX
@article{Farbin2026,
author = {Farbin, Redwana and Ismail, Sakurah and Paul, Subrata Kumer and Hamid, Md. Ekramul},
title = {Deep Learning Approaches for Hallucination Detection in Large Language Models: A Comprehensive Survey},
journal = {IOSR Journal of Computer Engineering},
year = {2026},
volume = {28},
number = {3},
pages = {18--30},
doi = {10.9790/0661-2803041830}
}
EUVIMED