Video evaluation
The project focused on evaluating AI model responses using reference video and structured prompt. For each task, two AI generated response were provided, and the goal was to select the response that best matched the reference video based on scene understanding, characters, background details, overall quality, and the presence of ZI errors.