Reasoning with LLMs

This one-week, intensive graduate-level course provides a deep technical and theoretical exploration of advanced reasoning with Large Language Models (LLMs). Moving beyond basic applications, students will investigate the fundamental mechanisms and state-of-the-art techniques that enable LLMs to perform complex, multi-step reasoning. The course covers neural reasoning methods, agentic reasoning with external tools, robust evaluation methodologies, and recent Reinforcement Learning based approaches and finally current open research questions. Through lectures, discussions, and a hands-on Python notebook, students will gain the skills to critically analyze, design, and implement reasoning systems.