Thaddäus Wiedemer
PhD Candidate
About Me
I’m a 4th-year PhD student in the International Max Planck Research School for Intelligent Systems (IMPRS-IS), currently interning at Google Deepmind Toronto. Most of my PhD focused on generalization, robustness, and benchmarking for large vision-language, language, and video models.
My current research investigates how generative video models can be used for multi-modal reasoning and how post-training can be leveraged for exploration beyond the pretraining data.
Education
PhD Machine Learning
Max Planck Institute for Intelligent Systems & University of Tübingen
M.Sc. Electrical Engineering and Information Technology
Karlsruhe Institute of Technology
B.Sc. Electrical Engineering and Information Technology
Karlsruhe Institute of Technology
Recent Publications
(2025).
VGGSounder: Audio-Visual Evaluations for Foundation Models.
ICCV 2025.
(2025).
LLMs on the Line: Data Determines Loss-to-Loss Scaling Laws.
ICML 2025.
(2024).
Pretraining Frequency Predicts Compositional Generalization of CLIP on Real-World Tasks.
NeurIPS 2024 Workshop.
(2024).
In Search of Forgotten Domain Generalization.
ICLR 2025 (Spotlight).
(2024).
Scale Learning in Scale-Equivariant Convolutional Networks.
VISAPP 2024.