AI & ML interests

Foundation Models, Scaling Laws, Continual Pretraining, Data-centric ML, Neuroscience, Human-Machine comparisons

Recent Activity

lukasthede  updated a dataset about 9 hours ago
bethgelab/MedKIT
lukasthede  published a dataset 2 days ago
bethgelab/MedKIT
View all activity

Organization Card

The official huggingface organisation for the Bethgelab, University of Tuebingen. We are an interdisciplinary group working on two broad themes:

  1. Foundation Models: agentic systems and benchmarking, recursive self-improvement, reinforcement learning for language models, scaling laws, continual pretraining, lifelong benchmarking and data-centric perspectives on (compositional) generalisation.
  2. Neuroscience: neural system identification, gaze prediction, object-centric motion perception, brain-inspired AI, human-machine comparisons