New research from Russia proposes an unconventional method to detect unrealistic AI-generated images – not by improving the accuracy of large vision-language models (LVLMs), but by intentionally ...