2 datasets found

The dataset used for training the InstructVid2Vid model, which consists of video-instruction-edited video triplets.

A novel benchmark for text rewriting with a focus on cross-sentence rewrites, covering a wide variety of rewriting types expressed through natural language instructions.