🧰 Vision-Language Models

The items that name it 3

• Detecting Adversarial Images through Response Profiles of Vision-Language Models
  paper2026-10-07It is written here as Vision-Language Models (VLMs).

  Frozen vision-language models whose predictions can be altered by adversarial perturbations while confidence and image-text similarity patterns appear plausible

• PromptFoo LM Security DB
  tool2026-07-18It is written here as vision-language models (unspecified).

  Evaluated for refusing unsafe/unanswerable image questions vs. correcting unsupported premises, per KoNA benchmark

• Decoy Images Amplify Caption-Mediated Defenses Against Encoded Jailbreaks
  paperIt is written here as Vision-Language Models (VLMs).

  Five frontier VLMs tested; also three open-weight VLMs served with no moderation layer