Aligning Large Language Models with Human Instructions, Preferences and Values

What it is about

You deal with problems and methods for aligning large language models (LLMs) to human instructions, preferences and values. After the seminar you will understand the goals, assessment methods and current research approaches of LLM alignment research and be able to discuss perspectives for improvements.

What you will be able to do

• Understand the basics and principles of LLM alignment
• Critically analyze research papers in the field of LLM alignment
• Reflect on ethical implications of LLM alignment
• Strengthen communication and discussion skills through presentations

What the module consists of

• SeminarLecture introduction by the instructor; student presentations and discussions on selected research papers

Teaching method

• Einführender VortragProviding fundamentals at the beginning of the seminar
• StudierendenpräsentationenDeepening individual topics through autonomous preparation of research papers
• Diskussionen im PlenumFostering critical thinking, feedback and exchange for deeper understanding