> The bounding box coordinates returned by GPT-4V did not match the position of the dog. I suppose it just doesn't take image dimensions into consideration, an

> The bounding box coordinates returned by GPT-4V did not match the position of the dog. I suppose it just doesn't take image dimensions into consideration, and needs to be provided with max dimensions, or prompted to give percentages or other absolute values instead of pixels.