Reproduce “Long Time No See: Benchmarking VLMs for Out-of-Sight Spatiotemporal Reasoning in Egocentric Videos” 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
Fangzhou Ma, Ivo Alexander Ban, Eren Homburg, Gabriele Goletto, Rémi Pautrat, Mahdi Rad, Chiara Plizzari, Marc Pollefeys
Beyond3D is a benchmark for testing whether models can track relocated objects in first-person kitchen videos after they leave view. It asks multiple-choice questions about visibility, movement history, fixture location, and 3D position, using video and text. Qwen-3.6-27B reaches 42.2% macro accuracy, above 29.7% chance, but this measures benchmark accuracy, not real-world retrieval.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.