作者位次 #6/7ACL 2026Workshop· Natural Language Processing
The Shape of Vulnerability: How Adversarial Perturbations Reshape the Topology of Language Model Latent Spaces
Angelina Tsai · Shreya Subramanian · Catherine Liu · Kimberly Lopez · Leif Zinn-Brooks · Alexia E. Schulz · Adaku Uchendu
摘要
Adversarial perturbations in the context of large language models (LLMs) are subtle changes added to input data (i.e., images or text) that are designed to alter predictions or outputs of machine learning models. We introduce several novel visualizations using topological data analysis (TDA) (lev...