سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / International Conference on Artificial Intelligence; City, Industry and Health

Personalizing Large Language Models: A Deep Dive into Adaptation Strategies

نویسندگان :

Negin Dehkhoda¹ Hamid Rastagari²

1- Department of Computer Engineering, Na.C., Islamic Azad University, Najafabad, Iran 2- Department of Computer Engineering, Na.C., Islamic Azad University, Najafabad, Iran

کلمات کلیدی :

Large Language Models (LLMs)،Retrieval-Augmented Generation،Parameter-Efficient Fine-Tuning،Personalization

چکیده :

Large Language Models (LLMs) have revolutionized Artificial Intelligence, with advanced capabilities across diverse applications. However, a significant challenge remains: how to personalize these models to meet specific user needs. Personalization is crucial in scenarios where tailored outputs are required, such as in industries where LLMs are deployed for tasks like customer service automation, personalized marketing, and decision support. In these contexts, the ability to adapt LLMs to specific user behaviors and organizational requirements can lead to better user engagement, increased operational efficiency, and enhanced decision-making. Researchers have explored two main strategies to achieve personalization: prompt-based methods and fine-tuning approaches. Prompt-based methods integrate user-specific content through dynamic prompts without modifying the underlying model parameters, offering cost-effective and flexible solutions. However, they struggle with capturing complex user behaviors over time. Fine-tuning, on the other hand, modifies the model’s parameters to better encode user-specific patterns, providing deeper customization at the cost of higher computational resources and data requirements. This review evaluates the strengths and limitations of both approaches and examines emerging techniques such as Retrieval-Augmented prompting and Parameter-Efficient Fine-Tuning, which aim to strike a balance between personalization, scalability, and industrial application.