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 actions. Unlike prior work that relies on prompt-based image generation and editing to perform action-conditioned image manipulation, our training hypothesis requires models to simulate the outcome of a real-world action and then reverse it to the original state. This forward-reverse requirement tests genuine cause-and-effect understanding rather than stylistic or semantic edits. We curate a high-quality benchmark of reversible actions from real-world scenarios to enable robust action grounding. Our experiments reveal that current models struggle with action reversibility, highlighting the need to evaluate action understanding. Do-Undo provides an intuitive testbed for evaluating and advancing action-aware generation in multimodal systems that must reason about real-world dynamics.
@article{arxiv.2512.13609,
title = {Do-Undo Bench: Reversibility for Action Understanding in Image Generation},
author = {Shweta Mahajan and Shreya Kadambi and Hoang Le and Rajeev Yasarla and Apratim Bhattacharyya and Munawar Hayat and Fatih Porikli},
journal= {arXiv preprint arXiv:2512.13609},
year = {2026}
}