Christian Schlarmann @chs20.bsky.social · 01/04/2025📜Paper: arxiv.org/abs/2502.11725arxiv.orgAdversarially Robust CLIP Models Can Induce Better (Robust) Perceptual MetricsMeasuring perceptual similarity is a key tool in computer vision. In recent years perceptual metrics based on features extracted from neural networks with large and diverse training sets, e.g. CLIP, h... 000
Christian Schlarmann @chs20.bsky.social · 01/04/2025📢 Robustness is not always at odds with accuracy! We show that adversarially robust vision encoders improve clean and robust accuracy over their base models in perceptual similarity tasks. Looking forward to presenting at SaTML @satml.org in Copenhagen next week 🇩🇰 131
Christian Schlarmann @chs20.bsky.social · 11/02/2025In line with previous works, this shows that it is important to develop adaptive attacks against new defenses in order to claim robustness. 📜 arxiv.org/abs/2411.14834arxiv.orgEvaluating the Robustness of the "Ensemble Everything Everywhere" DefenseEnsemble everything everywhere is a defense to adversarial examples that was recently proposed to make image classifiers robust. This defense works by ensembling a model's intermediate representations... 010
Christian Schlarmann @chs20.bsky.social · 11/02/2025📢 Check out our new report: we show that a recently proposed defense against adversarial attacks is not robust. We circumvent gradient masking issues of the proposed model by attacking a slightly adapted surrogate model and then transferring the perturbations. 151
Reposted by Christian SchlarmannTübingen AI Center @tuebingen-ai.bsky.social · 20/01/2025Great milestone for www.scholar-inbox.com! 🎊scholar-inbox.comScholar InboxScholar Inbox is a personal paper recommender which enables researchers to stay up-to-date with the most relevant progress in their field based on their personal research interests. Scholar Inbox is f... 092
