Reproduce “SCOPE: Scene-graph Construction Optimized by Plan Error-correction for Robot Long-term Planning Datasets” 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
Chaoyue Li, Zeyu Gao, Yanfeng Lu, Fan Luo
SCOPE generates diverse household scenes and instructions, then checks plans and revises scene graphs when a planner fails. In AI2-THOR simulation, its noisy-scene test achieved 70.67% task success, compared with DELTA’s 46.33%. Fine-tuning on SCOPE data achieved 63.00% success versus 50.67% with DELTA data; these results do not demonstrate physical-robot performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.