Detecting Contextual Hallucinations in Large Language Models with Frequency-Aware Attention

Siya Qi, Yudong Chen, Runcong Zhao, Qinglin Zhu, Zhanghao Hu, Wei Liu, Yulan He, Zheng Yuan, Lin Gui
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:100269-100299, 2026.

Abstract

Hallucination detection is critical for ensuring the reliability of large language models (LLMs) in context-based generation. Prior work has explored intrinsic signals available during generation, among which attention offers a direct view of grounding behavior. However, existing approaches typically rely on coarse summaries that fail to capture fine-grained instabilities in attention. Inspired by signal processing, we introduce a frequency-aware perspective on attention by analyzing its variation during generation. We model attention distributions as discrete signals and extract high-frequency components that reflect rapid local changes in attention. Our analysis reveals that hallucinated tokens are associated with high-frequency attention energy, reflecting fragmented and unstable grounding behavior. Based on this insight, we develop a lightweight hallucination detector using high-frequency attention features. Experiments on the RAGTruth and HalluRAG benchmarks show that our approach achieves performance gains over verification-based, internal-representation-based, and attention-based methods across models and tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-qi26d, title = {Detecting Contextual Hallucinations in Large Language Models with Frequency-Aware Attention}, author = {Qi, Siya and Chen, Yudong and Zhao, Runcong and Zhu, Qinglin and Hu, Zhanghao and Liu, Wei and He, Yulan and Yuan, Zheng and Gui, Lin}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {100269--100299}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/qi26d/qi26d.pdf}, url = {https://proceedings.mlr.press/v306/qi26d.html}, abstract = {Hallucination detection is critical for ensuring the reliability of large language models (LLMs) in context-based generation. Prior work has explored intrinsic signals available during generation, among which attention offers a direct view of grounding behavior. However, existing approaches typically rely on coarse summaries that fail to capture fine-grained instabilities in attention. Inspired by signal processing, we introduce a frequency-aware perspective on attention by analyzing its variation during generation. We model attention distributions as discrete signals and extract high-frequency components that reflect rapid local changes in attention. Our analysis reveals that hallucinated tokens are associated with high-frequency attention energy, reflecting fragmented and unstable grounding behavior. Based on this insight, we develop a lightweight hallucination detector using high-frequency attention features. Experiments on the RAGTruth and HalluRAG benchmarks show that our approach achieves performance gains over verification-based, internal-representation-based, and attention-based methods across models and tasks.} }
Endnote
%0 Conference Paper %T Detecting Contextual Hallucinations in Large Language Models with Frequency-Aware Attention %A Siya Qi %A Yudong Chen %A Runcong Zhao %A Qinglin Zhu %A Zhanghao Hu %A Wei Liu %A Yulan He %A Zheng Yuan %A Lin Gui %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-qi26d %I PMLR %P 100269--100299 %U https://proceedings.mlr.press/v306/qi26d.html %V 306 %X Hallucination detection is critical for ensuring the reliability of large language models (LLMs) in context-based generation. Prior work has explored intrinsic signals available during generation, among which attention offers a direct view of grounding behavior. However, existing approaches typically rely on coarse summaries that fail to capture fine-grained instabilities in attention. Inspired by signal processing, we introduce a frequency-aware perspective on attention by analyzing its variation during generation. We model attention distributions as discrete signals and extract high-frequency components that reflect rapid local changes in attention. Our analysis reveals that hallucinated tokens are associated with high-frequency attention energy, reflecting fragmented and unstable grounding behavior. Based on this insight, we develop a lightweight hallucination detector using high-frequency attention features. Experiments on the RAGTruth and HalluRAG benchmarks show that our approach achieves performance gains over verification-based, internal-representation-based, and attention-based methods across models and tasks.
APA
Qi, S., Chen, Y., Zhao, R., Zhu, Q., Hu, Z., Liu, W., He, Y., Yuan, Z. & Gui, L.. (2026). Detecting Contextual Hallucinations in Large Language Models with Frequency-Aware Attention. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:100269-100299 Available from https://proceedings.mlr.press/v306/qi26d.html.

Related Material