Rates the perceptual quality or aesthetics of an image using the Qwen3.5-VL multimodal model. The result is a score ranging from 0 (worst) to 1 (best).