New arXiv work introduces Prompt-Adapted adversarial images for vision-language models, selectively rejecting harmful prompts while still answering legitimate ones. A useful step toward more nuanced visual privacy…

#AIresearch #ComputerVision #VLMs #AdversarialML
https://arxiv.org/abs/2610.16622

New arXiv work characterizes how small, coordinated changes to comparison data can sharply destabilize maximum-likelihood pairwise ranking, framing the issue as a budgeted subset-selection problem and proposing…

#MachineLearning #Ranking #DataScience #AdversarialML
https://arxiv.org/abs/2604.17805

Split Computing for DNN inference faces combined hardware fault and adversarial risks, with a new paper proposing the Joint Vulnerability Score to assess both together. Relevant for edge AI device reliability.

#SplitComputing #EdgeAI #AdversarialML #DNNRobustness
https://arxiv.org/abs/2610.04396

arXiv 2609.33898 reports the first systematic cross-domain study showing efficiency-oriented training of vision and language models increases susceptibility to adversarial and privacy attacks.

#AIResearch #ModelRobustness #AdversarialML
https://arxiv.org/abs/2609.33898