Hung Huy Nguyen

SketchVLM: Vision language models can annotate images to explain thoughts and guide users

SketchVLM is presented, a training-free, model-agnostic framework that enables VLMs to produce non-destructive, editable SVG overlays on the input image to visually explain their answers, and finds that single-turn generation already achieves strong accuracy and annotation quality, and multi-turn generation opens up fu...