Claas Beger

Claas is interested in building AI systems that can reason like humans, not only in outcome, but in the underlying thought process. For this purpose, he is interested in varying perspectives on natural intelligence, including principles of neuroscience, human cognition and the overall processes of learning. In the past, he has done projects on a Spiking Neural Network Transformer, a Hippocampus-inspired architecture for deriving and applying rules in compositional statements, Evaluating and Solving Bongard Puzzles and Structured Agentic workflows.

Claas recently finished a Master's degree in Computer Science at Cornell. At SFI, he is working with Melanie Mitchell on evaluating and advancing multimodal reasoning capabilities in large Vision-Language Models. This includes the development of new benchmarks and neurosymbolic systems that draw on prior knowledge and visual routines to enable more interpretable, compositional behavior.

Prior to Cornell, Claas earned his bachelor’s degree from the Technical University of Munich with a semester abroad at The Hong Kong University of Science and Technology. Outside of AI research, he enjoys playing (Beach-)Volleyball, Basketball and working out, as well as watching movies.