Hugging Face Daily Papers·· 12 days agoAI score44
MotionGraphicsBench: Evaluating Motion Graphics Generation with Programmatic Verification
AI briefing
The MotionGraphicsBench benchmark contains 100 instances and is used to evaluate large language models' ability to generate motion graphics from static SVG and motion editing prompts. On a balanced test set of 180 animations, its programmatic verification achieved a raw agreement rate of 93.3% with human judgment, higher than the GPT-6 Astra VLM baseline's 83.3%.
Open original arxiv.org