1 paper
Shweta Mahajan, Shreya Kadambi, Hoang Le +4
We introduce the Do-Undo task and benchmark to address a critical gap in vision-language models: understanding and generating plausible scene transformations driven by real-world a…