Innovative approaches to optimization of large language models

Modern methods and approaches to improving large language models (LLM), including architecture optimization, improving the quality of training data, introducing new learning strategies and taking into account ethical aspects are considered. The theoretical foundations of each method are given, accompanied by examples of using modern technologies and models GPT-4, Llama 3.2 and Claude 3.5, which allows for a better understanding of their practical application and significance are given. Keywords large language models, approaches, methods, architecture optimization