AI vision hallucinations caught out by new tool, says Los Alamos
The Prelim Attention Score tool helps detect whether a model’s output is grounded in the image or driven too strongly by its own generated text. (Image: Los Alamos National Laboratory)
The Prelim Attention Score system improves the trustworthiness of computer vision-based AI models, says the National Laboratory
Register for FREE to keep reading
Join 10,000+ vision professionals driving innovation in automation, AI and imaging with:
- Expert insights on vision, robotics, AI & embedded tech
- Newsletters and features covering the full imaging landscape
- Visionaries series: leadership strategies in imaging
- Free panels on smart manufacturing & autonomy
- White Papers & updates for smarter integration
Sign up now
Already a member? Log in here
Your data is protected under our privacy policy.
