Reproduce “The Prestige: Benchmarking Cognitive Visual Reasoning using Magic Tricks” 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
Shuo Wen, Beixi Du, Edwin Meriaux, Junming Shi, Chloe Si, David Meger, Doina Precup, Gregory Dudek
The authors introduce The Prestige, a benchmark testing whether video-language models can explain professional magic performances through open-ended questions. It contains 180 videos and 2,177 questions across six abilities, emphasizing contextual and paradoxical reasoning. Gemini3.1-Pro scored 47.08% overall, compared with 92.79% for human reviewers, on this benchmark rather than video understanding generally.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.