Doctoral Thesis Defense of Kyra Ahrens

18 July 2025

On 18.07.2025, our colleague Kyra Ahrens had the successful defense of her doctoral thesis "Methods for Downstream Continual Learning with Foundation Models". We congratulate her for successful graduation and wishing the best for her future career! Here is brief information about her thesis:

Continual learning addresses the challenge of training deep neural networks on a sequence of tasks while avoiding catastrophic forgetting. This thesis addresses shortcomings in downstream continual fine-tuning of foundation models by introducing and evaluating four different strategies: dual-memory replay, multi-layer prototyping, selective specialization, and noise-augmented feature replay, each delivering strong performance across diverse unimodal classification and multimodal reasoning problems.

Here are some highlight publications that arose during her doctorate: 

Visually Grounded Continual Language Learning with Selective Specialization 

Read Between the Layers: Leveraging Multi-Layer Representations for Rehearsal-Free Continual Learning with Pre-Trained Models 

Unconstrained Open Vocabulary Image Classification: Zero-Shot Transfer from Text to Image via CLIP Inversion