Large Language Models and Agents (HWS2024)
Large language models (LLMs) such as GPT, Llama, Gemini, and Mixtral have the potential to enable a wide range of new applications and to significantly improve the performance of existing systems. The course introduces students to LLMs and teaches them how to employ the models within applications.
The course covers the following topics:
- Introduction to LLMs
- Instruction Tuning and Reinforcement Learning from Human Feedback
- Prompt Engineering and Efficient Adaptation
- LLM-based Agents
- Evaluation of LLMs and Agents
- Development of LLM-based Applications
The course participants gain knowledge about the principles of training LLMs. They will be able to identify opportunities for employing LLMs in business applications and will learn to apply prompt engineering techniques as well as agent frameworks for solving complex tasks. In the second half of the course, the participants apply their knowledge in team projects and will report about the results of the projects in the form of a written report as well as an oral presentation.
This course will again be offered in FSS2025
Due to high demand, this course will again be offered in FSS2025. The lectures will again take place Thursdays 15:30 to 17:00. Further details will follow.
Time and Location
- Thursdays, 15:30–17:00. Location: A5 C015 (Starting: 12.09.2024)
- Attention: Exercise in first October week will be held on Wednesday 02.10. instead of Thursday. Room B6 A1.01 from 17:15–18:45.
Instructors
Assessment
In the second half of the course, students will work on applied projects in teams. The grade received for this course is based on the quality of the contents of their project report (70%) and the associated presentation (30%).
There is no dedicated written or oral exam for this course.
ECTS
- 3 ECTS: 70 % project report, 30 % presentation
Requirements
- Programming skills in Python
- Basic machine learning concepts and techniques
Registration and Participation
- The course is open to students of the Mannheim Master in Data Science, the Mannheim Master in Social Data Science, and Master Business Informatics.
- The course is restricted to 30 participants. The registration for the course is organized by the Studiengang Management and is done via Portal2.
Outline and Course Material
Day | Topic | Additional Material |
---|---|---|
05.09.2024 | - no lecture – | - |
12.09.2024 | Lecture: Introduction to Language Models | Tutorials/ |
19.09.2024 | Lecture: Instruction Tuning and RLHF | |
26.09.2024 | Lecture: Prompt Engineering and Efficient Adaptation | |
02.10.2024 | Exercise:Introduction to LangChain | Solution |
10.10.2024 | Lecture: LLM Agents and Tool Use | |
17.10.2024 | Exercise: Introduction to LangGraph | Solution/Project Topics |
24.10.2024 | Project: Introduction to Student Projects | |
31.10.2024 | Project Coaching | |
07.11.2024 | Project Coaching | |
14.11.2024 | Project Coaching | |
21.11.2024 | Project Coaching | |
28.11.2024 | Project Coaching | |
05.12.2024 | Presentation of Project Results |
Literature
- Daniel Jurafsky & James H. Martin: Speech and Language Processing.
- Zhao et al.: A Survey of Large Language Models. 2024. arXiv:2303.18223
- Wang et al.: A Survey on Large Language Model based Autonomous Agents. 2024. arXiv:2302.07842
- Zhou et al.: A Comprehensive Survey on Pretrained Foundation Models: A History from BERT to ChatGPT. 2023. arXiv:2302.09419.