Description

Review AI-generated image captions in English and flag where they do not match the image.

You will work through a fixed set of images, reading each caption in full against the image and logging every defect you can substantiate — wrong objects or counts, wrong colours, misread text rendered inside the image, claims the image does not support. The image is the ground truth, and careful reading matters more than speed.

Requirements: native-level English, strong attention to visual and textual detail, and the patience to check a caption line by line against what is actually in front of you. No specialist degree required.