iiinova — where your digital self keeps evolving

Reposted by Andi Zhang

Gen AI - the AI Hub in Generative Models @genaihub.bsky.social · 04/03/2026

A method for testing AI safety by using human-like concepts to trick generative models into making mistakes, has been developed by Hub researchers. The paper, entitled Concept-based Adversarial Attack: A Probabilistic Perspective, is summarised on our website: www.genai.ac.uk/news/hub-res...

A series of images included in the paper including dogs and rubber ducks, providing a qualitative comparison. A green border indicates an example that successfully fools the classifier; red indicates failure.

011