Understanding the basics of large language models
Mehrsa Pourya, Doctoral Student at EPFL

Seminar • 2025-10-21

Abstract
We are going to review different steps of large language models: tokenization, architecture, and training loss. Large Language Models (LLMs) have transformed natural language processing by using transformer architectures and large-scale datasets. This presentation provides a concise overview of how text is represented through tokenization, how the model processes information via its architecture, and how it learns patterns using training objectives.