Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{208736,
        author = {Sai Gadhave and Krutika Sonare and Anushka Varpe and Anushka Waghchoure and Aaryan Laygude and Dr. Shital Ghotekar},
        title = {Intent Hallucination in Large Language Models: A Systematic Review of Instruction Following, Detection, And Mitigation Techniques},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {no},
        pages = {614-620},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=208736},
        abstract = {Large Language Models (LLMs) have demonstrated significant capabilities in generating human-like and contextually relevant responses across a wide range of applications. However, their reliability is affected not only by factual hallucinations but also by failures to correctly understand and follow user instructions. This emerging issue, referred to as intent hallucination, occurs when an LLM omits, misinterprets, or fails to satisfy one or more requirements contained in a user's request, even when the generated response may be factually correct. Such failures can reduce the reliability and usefulness of LLM-based systems, particularly in complex tasks involving multiple constraints. 
This paper proposes a systematic review of existing research on intent hallucination and instruction-following capabilities in Large Language Models. The study will examine major types and causes of instruction-following failures, existing benchmarks and evaluation techniques, methods for detecting such failures, and approaches proposed to mitigate them. The review will also compare current evaluation frameworks and investigate limitations in existing research, particularly concerning complex, multi-constraint, and dynamic user instructions. Based on the reviewed literature, the study aims to organize existing approaches into a structured taxonomy covering instruction-following failures, detection and evaluation methods, and mitigation strategies. Finally, the paper will highlight open research challenges and future directions for developing more reliable and intent-aligned LLM systems. The study aims to provide a consolidated understanding of intent hallucination and support future research toward more dependable human–AI interaction.},
        keywords = {Large Language Models, Generative AI, Intent Hallucination, Instruction Following, LLM Evaluation, Hallucination Detection, Prompt Engineering, Artificial Intelligence},
        month = {September},
        }