3 papers
cs.CR2025
ShadowLogic: Backdoors in Any Whitebox LLM
Kasimir Schulz, Amelia Kawasaki, Leo Ring
Large language models (LLMs) are widely deployed across various applications, often with safeguards to prevent the generation of harmful or restricted content. However, these safeg…
cs.LG2025
TokenBreak: Bypassing Text Classification Models Through Token Manipulation
Kasimir Schulz, Kenneth Yeung, Kieran Evans
Natural Language Processing (NLP) models are used for text-related tasks such as classification and generation. To complete these tasks, input data is first tokenized from human-re…
cs.LG2025
ShadowGenes: Leveraging Recurring Patterns within Computational Graphs for Model Genealogy
Kasimir Schulz, Kieran Evans
Machine learning model genealogy enables practitioners to determine which architectural family a neural network belongs to. In this paper, we introduce ShadowGenes, a novel, signat…