Reproduce “Generalizable AI-Generated Image Detection with Embedding-Space Explanations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Nikos Giakoumoglou, Paschalis Giakoumoglou, Tania Stathaki, Symeon Papadopoulos
FOX detects whether an image is real or fake and generates an explanation without allowing explanation training to change its verdict. It adapts a visual detector, locks it, and passes detector-derived evidence through learned queries to a language model. Across 14 unseen generators, FOX achieves mean AUC 97.4, 3.7 points above the runner-up; this is whole-image detection, not localization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.