Seminar

How formal frameworks from computer science and philosophy, such as temporal logics, reward machines, deontic logics and theories of reasons, can make AI alignment more rigorous and interpretable.

The most important deep reinforcement learning algorithms, followed by special topics from our research: exploration, AI planning, and agentic reinforcement learning, with a practical project in teams.

Statistical and machine learning models for data that evolves over space and time, from Gaussian and point processes to Bayesian neural fields and spatiotemporal graph neural networks.

The basics of graph neural networks and their applications in various domains, summarized by the participants in easy-to-follow talktorials.

Neural network-based policies that solve all instances of a planning domain, built on graph neural networks or transformers.

Block seminar on large language models, from the foundations of language representations and transformers to integrating LLMs with neuro-symbolic reasoning, knowledge graphs and other structured data.

The most important deep reinforcement learning algorithms, followed by special topics from our research: exploration, safe reinforcement learning, and causal reinforcement learning.

Advanced diffusion-model topics including discrete diffusion, equivariant models, consistency models, and reinforcement-learning-based training, paired with practical tutorial notebooks.

Project-based seminar where teams build AI solutions to real-world social problems. The 2023 edition produced Safe Streets, Disaster Tweet, and Simple English.

Theoretical foundations of diffusion models and applications across image, video, molecular, and graph generation.