Reproduce “Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographies” 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
Hui Ren, Lei Fan, Henry Pao, Han Guo, Zeeshan Zia, Ying Chen, Alexander Schwing, Gang Hua
GEB records each physical instance’s visually grounded encounters, links them cautiously, and retains episodes and source videos. It uses these biographies to connect moments and retrieve context across long videos. Under matched controller and retrieval limits, EgoLifeQA accuracy is 72.0% versus MAGIC-Video’s 67.6%, but identity is estimated, not guaranteed, and retrieval is slower than caption-only methods.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.