VISUAL PROMPTS IN VIDEO MODELS

Summary from the TLDR AI newsletter

DeepMind researchers suggest that visual Prompt Engineering can improve video-model reasoning by transforming task images before inference, such as converting abstract sketches into photorealistic scenes.