Profile

Lukas is currently pursuing his PhD under joined supervision of Zeynep Akata (at Helmholtz Munich and TUM) and Matthias Bethge (at the University of Tübingen) as part of the International Max-Planck Research School for Intelligent Systems (IMPRS-IS). During his PhD, he completed a research internship at Thomson Reuters Foundational Research. He obtained his bachelor's and master's degrees in Industrial Engineering and Management from the Karlsruhe Institute of Technology (KIT). While studying, Lukas completed internships at Mercedes-Benz in Stuttgart and at Ernst & Young in Berlin.

His research focuses on the reliable and efficient adaptation of foundation models. He is interested in how large pretrained models can be adapted to new domains, tasks, and evolving knowledge while preserving the capabilities that make them useful in the first place. His current work spans three complementary directions: leveraging the internal structure of pretrained models for efficient adaptation, continually adapting foundation models to evolving knowledge, and understanding adaptation dynamics and capability drift across both vision and language models.

Publications

"CapTrack: Multifaceted Evaluation of Forgetting in LLM Post-Training" Lukas Thede, Stefan Winzeck, Zeynep Akata, Jonathan Richard Schwarz

"Structural Pruning of Large Vision Language Models: A Comprehensive Study on Pruning Dynamics, Recovery, and Data Efficiency" Yiran Huang, Lukas Thede, Massimiliano Mancini, Wenjia Xu, Zeynep Akata International Journal of Computer Vision, IJCV 2026

"Beyond the final layer: Attentive multilayer fusion for vision transformers" Laure Ciernik*, Marco Morik*, Lukas Thede, Luca Eyring, Shinichi Nakajima, Zeynep Akata, Lukas Muttenthaler International Conference on Machine Learning, ICML 2026

"Understanding the Limits of Lifelong Knowledge Editing in LLMs" Lukas Thede, Karsten Roth, Matthias Bethge, Zeynep Akata, Tom Hartvigsen International Conference on Machine Learning, ICML 2025

"Reflecting on the State of Rehearsal-free Continual Learning with Pretrained Models"
Lukas Thede*, Karsten Roth*, Olivier J. Hénaff, Matthias Bethge, Zeynep Akata
Conference on Lifelong Learning Agents, CoLLAs 2024

"Fantastic Gains and Where to Find Them: On the Existence and Prospect of General Knowledge Transfer between Any Pretrained Model"
Karsten Roth*, Lukas Thede*, A. Sophia Koepke, Oriol Vinyals, Olivier Henaff, Zeynep Akata
Spotlight at International Conference on Learning Representations, ICLR 2024