Textual Intelligence

9Economizing Large Language Model Training and Alignment with Human Values through Cost Effective Architectures and Transfer Learning Techniques

Mohammed Wasim Bhatt¹*, Rubal Jeet², Mukesh Soni³, Haewon Byeon⁴ and Vishal Sagar⁵

¹Model Institute of Engineering and Technology, Jammu, J&K, India

²Department of Computer Science and Engineering, Chandigarh Engineering College, Chandigarh Group of Colleges, Punjab, India

³Dr. D. Y. Patil Vidyapeeth, Pune, Dr. D. Y. Patil School of Science & Technology, Tathawade, Pune, India

⁴Worker’s Care & Digital Health Lab, Department of Future Technology, Korea University of Technology and Education, South Korea

⁵Graphic Era Hill University, Graphic Era Deemed to be University, Dehradun, India

Abstract

This research looks at ways to train large language models (LLMs) at a minimal cost while maintaining their human values using cost-effective designs and transfer learning. A lack of ethical consistency and high computing costs have hindered LLM progress. The study’s main emphasis is on finding answers to each problem. Even if it improves AI, learning LLMs may pose ethical concerns and increase processing costs. Ethically using these concepts necessitates aligning them with human rights and values. The study’s goal is to uncover novel strategies to reduce training costs while maintaining LLM integrity and quality. The proposed technique relies heavily on cost-effective architectural designs and transfer learning methods. We examine transfer learning, ...