This paper was accepted at the Principled Design for Trustworthy AI — Interpretability, Robustness, and Safety across Modalities Workshop at ICLR 2026.
What exactly makes a particular image unsafe? Systematically differentiating between benign and problematic images is a challenging problem, as subtle changes to an image, such as an insulting gesture or symbol, can drastically alter its safety implications. However, existing image safety datasets are coarse and ambiguous, offering only broad safety labels without isolating the specific features that drive these differences. We introduce… Read More
SafetyPairs: Isolating Safety Critical Image Features with Counterfactual Image Generation
This paper was accepted at the Principled Design for Trustworthy AI — Interpretability, Robustness, and Safety across Modalities Workshop at ICLR 2026.
What exactly makes a particular image unsafe? Systematically differentiating between benign and problematic images is a challenging problem, as subtle changes to an image, such as an insulting gesture or symbol, can drastically alter its safety implications. However, existing image safety datasets are coarse and ambiguous, offering only broad safety labels without isolating the specific features that drive these differences. We introduce…
Related Posts
4 highlights from Google Beam in 2025
Google Beam, our first true-to-life 3D video communication platform, made great progress in 2025.
The Potential of CoT for Reasoning: A Closer Look at Trace Dynamics
Chain-of-thought (CoT) prompting is a de-facto standard technique to elicit reasoning-like responses from large language…
At our Research@ Poland event we shared how AI is helping us solve big challenges.
From AI education to disaster response, see how collaboration is at the heart of the…