Reproduce “Error-Driven Scene Editing for 3D Grounding in Large Language Models” 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
Yue Zhang, Zun Wang, Han Lin, Jialu Li, Jianing Yang, Yonatan Bitton, Idan Szpektor, Mohit Bansal
DEER-3D targets 3D grounding errors by diagnosing predicates a baseline gets wrong, then editing existing scenes to create counterfactual examples and aligned questions for training. It repeats this process across rounds. On ScanRefer, Chat-Scene’s [email protected] rises from 50.2% to 53.3% after Round 2, a 3.1-point gain; this benchmark result does not show errors are generally eliminated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.