Text to Image H2H
Performed Head-to-Head (H2H) evaluations for text-to-image generative models, side-grading outputs based on semantic alignment (how well the image matched the prompt) and visual fidelity (resolution, lighting, and absence of artifacts)