Large Language Models: A Structured Taxonomy and Review of Challenges, Limitations, Solutions, and Future Directions
Большие языковые модели: структурированная таксономия и обзор проблем, ограничений, решений и перспективных направлений
2025-07-21
SCID: 54.1/8h3vef2u
Discuss with AI
algorithmic biasdata privacy and securitylarge language modelsmodel interpretabilitynatural language processing
Figures from the paper
Abstract (AI)
Large language models (LLMs), as one of the most advanced achievements in the field of natural language processing (NLP), have made significant progress in areas such as natural language understanding and generation. However, attempts to achieve the widespread use of these models have met numerous challenges, encompassing technical, social, ethical, and legal aspects. This paper provides a comprehensive review of the various challenges associated with LLMs and analyzes the key issues related to these technologies. Among the challenges discussed are model interpretability, biases in data and model outcomes, ethical concerns regarding privacy and data security, and their high computational requirements. Furthermore, the paper examines how these challenges impact the applications of LLMs in fields such as healthcare, law, media, and education, emphasizing the importance of addressing these issues in the development and deployment of these models. Additionally, solutions for improving the robustness and control of models against biases and quality issues are proposed. Finally, the paper looks at the future of LLM research and the challenges that need to be addressed for the responsible and effective use of this technology. The goal of this paper is to provide a comprehensive analysis of the challenges and issues surrounding LLMs in order to enable the optimal and ethical use of these technologies in real-world applications.
Key Findings
1
LLMs have advanced natural-language understanding and generation but face technical, social, ethical, and legal barriers to widespread adoption.
2
Major challenges include limited interpretability, biased training data and outputs, privacy and data-security risks, and high computational requirements.
3
Responsible and effective real-world use of LLMs requires continued research addressing their technical limitations and broader societal risks.
4
The review discusses solutions for improving model robustness and controlling bias and quality-related failures.
5
These challenges directly affect LLM deployment in healthcare, law, media, and education, requiring domain-sensitive safeguards.
Research Object
large language models (LLMs)
Research Subject
technical, social, ethical, and legal challenges, limitations, solutions, and future directions for the responsible and effective deployment of LLMs
Publication Details
Publication Date
2025-07-21
Journal
Publisher
ISSN
Cited by
67
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai16
"Why Should I Trust You?": Explaining the Predictions of Any Classifier2016
A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures2019
A Survey on Evaluation of Large Language Models2024
Recurrent neural network based language model2010
Training language models to follow instructions with human feedback2022
A Comprehensive Overview of Large Language Models2025
PaLM: Scaling Language Modeling with Pathways2022
BloombergGPT: A Large Language Model for Finance2023
Practical and ethical challenges of large language models in education: A systematic scoping review2023
Flamingo: a Visual Language Model for Few-Shot Learning2022
A Review on Large Language Models: Architectures, Applications, Taxonomies, Open Issues and Challenges2024
Transformer in Transformer2021
Can Large Language Models Transform Computational Social Science?2023
MM-LLMs: Recent Advances in MultiModal Large Language Models2024
A Review of Current Trends, Techniques, and Challenges in Large Language Models (LLMs)2024
Large Language Models: A Comprehensive Survey of its Applications, Challenges, Limitations, and Future Prospects2023