392144 Agentic AI: Architectures, Self-Evolution, and Safety (S) (SoSe 2026)

This Master's seminar explores the frontier of Advanced Intelligent Agents, investigating the challenges of moving beyond Large Language Models (LLMs) towards fully integrated, autonomous systems. The seminar focuses on modular system design, where cognitive, perceptual, and operational components are systematically mapped and integrated. We will critically investigate the core computational modules of advanced agents (including sophisticated memory, world modeling, reward processing, and goal management) and examine the mechanisms of agent self-evolution and continuous learning in dynamic environments. A central theme is the imperative of safe and beneficial AI, covering ethical alignment, robustness, and practical mitigation strategies necessary for trustworthy real-world deployment. This course is anchored in the comprehensive review "Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems"