Picture for Tom Wollschläger

Tom Wollschläger

REINFORCE Adversarial Attacks on Large Language Models: An Adaptive, Distributional, and Semantic Objective

Add code
Feb 24, 2025
Viaarxiv icon

The Geometry of Refusal in Large Language Models: Concept Cones and Representational Independence

Add code
Feb 24, 2025
Viaarxiv icon

Adversarial Alignment for LLMs Requires Simpler, Reproducible, and More Measurable Objectives

Add code
Feb 17, 2025
Viaarxiv icon

Certifiably Robust Encoding Schemes

Add code
Aug 02, 2024
Figure 1 for Certifiably Robust Encoding Schemes
Figure 2 for Certifiably Robust Encoding Schemes
Figure 3 for Certifiably Robust Encoding Schemes
Figure 4 for Certifiably Robust Encoding Schemes
Viaarxiv icon

Discrete Randomized Smoothing Meets Quantum Computing

Add code
Aug 01, 2024
Viaarxiv icon

Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space

Add code
Jun 15, 2024
Figure 1 for Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space
Figure 2 for Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space
Figure 3 for Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space
Figure 4 for Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space
Viaarxiv icon

Expressivity and Generalization: Fragment-Biases for Molecular GNNs

Add code
Jun 12, 2024
Viaarxiv icon

Energy-based Epistemic Uncertainty for Graph Neural Networks

Add code
Jun 06, 2024
Viaarxiv icon

Uncertainty for Active Learning on Graphs

Add code
May 02, 2024
Viaarxiv icon

Attacking Large Language Models with Projected Gradient Descent

Add code
Feb 14, 2024
Viaarxiv icon