Hallucination in Generative AI-Enabled Wireless Communications: Concepts, Issues, and Solutions
Authors:
Top Cited Papers
- 12 March 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Magazine
- Vol. 64 (5) , 164-171
Abstract
Generative AI (GenAI) is driving the intelligence of wireless communications. Due to data limitations, random generation, and dynamic environments, GenAI may generate channel information or optimization strategies that violate physical laws or deviate from actual real-world requirements. We refer to this phenomenon as hallucination in GenAI-enabled wireless communications, which results in invalid channel information, spectrum wastage, and low communication reliability but remains underexplored. To address this gap, this article provides a comprehensive concept of wireless hallucinations in GenAI-driven communications, focusing on hallucination mitigation. Specifically, we first introduce the fundamental, analyze its causes based on the GenAI workflow, and propose mitigation solutions at the data, model, and post-generation levels. Then, we systematically examine representative hallucination scenarios in GenAI-enabled communications and their corresponding solutions. Finally, we propose a novel integrated mitigation solution for GenAIbased channel estimation. At the data level, we establish a channel estimation hallucination dataset and employ generative adversarial networks (GANs)-based data augmentation. Additionally, we incorporate attention mechanisms and large language models (LLMs) to enhance both training and inference performance. Experimental results demonstrate that the proposed hybrid solutions reduce the normalized mean square error (NMSE) by 0.19, effectively reducing hallucinations in wireless communications.Keywords
Funding Information
- National Natural Science Foundation of China (62571057)
- Natural Science Foundation of Beijing Municipality (L254061)
- Beijing University of Posts and Telecommunications (2025TSQY03)
- Nanyang Technological University
This publication has 10 references indexed in Scilit:
- Generative AI Enabled Matching for 6G Multiple AccessIEEE Wireless Communications, 2025
- Hallucination Detection in Foundation Models for Decision-Making: A Flexible Definition and Review of the State of the ArtACM Computing Surveys, 2025
- Intelligent Reflecting Surface-Assisted Millimeter Wave Communications: Cross Attention-Aided Variational Autoencoder-Based Precoding DesignIEEE Transactions on Cognitive Communications and Networking, 2025
- A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open QuestionsACM Transactions on Information Systems, 2025
- Enhancing Physical Layer Communication Security Through Generative AI with Mixture of ExpertsIEEE Wireless Communications, 2025
- Generative AI on SpectrumNet : An Open Benchmark of Multiband 3-D Radio MapsIEEE Transactions on Cognitive Communications and Networking, 2024
- RF-Diffusion: Radio Signal Generation via Time-Frequency Diffusion2024
- Enhancing Reasoning Ability in Semantic Communication Through Generative AI-Assisted Knowledge ConstructionIEEE Communications Letters, 2024
- A Flexible Encoding/Decoding Procedure for 6G SCMA Wireless Networks via Adversarial Machine Learning TechniquesIEEE Transactions on Vehicular Technology, 2022
- Image-to-Image Translation with Conditional Adversarial Networks2017