Detecting Hallucination in Large Language Models Through Deep Internal Representation Analysis

Zhang, Luan; Song, Dandan; Wu, Zhijing; Tian, Yuhang; Zhou, Changzhi; Xu, Jing; Yang, Ziyi; Zhang, Shuhao

doi:10.24963/ijcai.2025/929

Detecting Hallucination in Large Language Models Through Deep Internal Representation Analysis

Luan Zhang, Dandan Song, Zhijing Wu, Yuhang Tian, Changzhi Zhou, Jing Xu, Ziyi Yang, Shuhao Zhang

Download BibTeX Cite

Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence. Main Track. Pages 8357-8365.

https://doi.org/10.24963/ijcai.2025/929

Keywords

Natural Language Processing: NLP: Resources and evaluation
Natural Language Processing: NLP: Question answering
Natural Language Processing: NLP: Language models
Natural Language Processing: NLP: Language generation

Abstract

Large language models (LLMs) have shown exceptional performance across various domains. However, LLMs are prone to hallucinate facts and generate non-factual responses, which can undermine their reliability in real-world applications. Current hallucination detection methods suffer from external resource demands, substantial time overhead, difficulty overcoming LLMs' intrinsic limitation, and insufficient modeling. In this paper, we propose MHAD, a novel internal-representation-based hallucination detection method. MHAD utilizes linear probing to select neurons and layers within LLMs. The selected neurons and layers are demonstrated with significant awareness of hallucinations at the initial and final generation steps. By concatenating the outputs from these selected neurons of selected layers at the initial and final generation steps, a hallucination awareness vector is formed, enabling precise hallucination detection via an MLP. Additionally, we introduce SOQHD, a novel benchmark for evaluating hallucination detection in Open-Domain QA (ODQA). Extensive experiments show that MHAD outperforms existing hallucination detection methods across multiple LLMs, demonstrating superior effectiveness.

Luan Zhang, Dandan Song, Zhijing Wu, Yuhang Tian, Changzhi Zhou, Jing Xu, Ziyi Yang, and Shuhao Zhang. 2025. Detecting Hallucination in Large Language Models Through Deep Internal Representation Analysis. In Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25. IJCAI, 8357–8365. https://doi.org/10.24963/ijcai.2025/929

@inproceedings{ijcai2025p929,
  title     = {Detecting Hallucination in Large Language Models Through Deep Internal Representation Analysis},
  author    = {Zhang, Luan and Song, Dandan and Wu, Zhijing and Tian, Yuhang and Zhou, Changzhi and Xu, Jing and Yang, Ziyi and Zhang, Shuhao},
  booktitle = {Proceedings of the Thirty-Fourth International Joint Conference on
               Artificial Intelligence, {IJCAI-25}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {James Kwok},
  pages     = {8357--8365},
  year      = {2025},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2025/929},
  url       = {https://doi.org/10.24963/ijcai.2025/929},
}

TY  - CPAPER
TI  - Detecting Hallucination in Large Language Models Through Deep Internal Representation Analysis
AU  - Zhang, Luan
AU  - Song, Dandan
AU  - Wu, Zhijing
AU  - Tian, Yuhang
AU  - Zhou, Changzhi
AU  - Xu, Jing
AU  - Yang, Ziyi
AU  - Zhang, Shuhao
T2  - Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2025
SP  - 8357
EP  - 8365
DO  - 10.24963/ijcai.2025/929
UR  - https://doi.org/10.24963/ijcai.2025/929
ER  -