The Adaptive Stress Testing for Robust AI (ASTRA) toolbox provides tooling to support model developers and testing in the full life cycle of making more robust AI Systems through the application of adaptive stress testing and adversarial training.
(2021) Robust Deepfake Detection project for the Deep Learning course at ETH. Authors: David Kamm, Nicolas Muntwyler, Alexander Timans, Moritz Vandenhirtz
Research collaboration conducted from the Palace of Science (Belgrade) in collaboration with Prof. El Mahdi El Mhamdi (École Polytechnique, Paris). This collaboration originated in the context of a French Government Scholarship (BGF – Bourse du Gouvernement français) awarded for PhD cotutelle studies
Framework using UMAP-DBSCAN for unsupervised discovery of multi-modal Hidden Bias Subgroups (HBSs) in AI failure spaces. Implements a scalable Multi-Domain MMD Objective to mitigate latent Acquisition Bias and enhance robustness in clinical Ocular Disease Recognition (ODR).
Learned influence-topology control for robust multi-agent LLM networks.
A hierarchical AI safety architecture with asymmetric supervisory control.
'Robust Deepfake Detection' project for the Deep Learning course at ETH Zurich, 2021. Authors (alphabetic): David Kamm, Nicolas Muntwyler, Alexander Timans, Moritz Vandenhirtz.
Repository with the project of the Explainable and Reliable Artificial Intelligence course at UniTS (2024-2025).
Add a description, image, and links to the robust-ai topic page so that developers can more easily learn about it.
To associate your repository with the robust-ai topic, visit your repo's landing page and select "manage topics."