Reproduce “LAGR: Language-Aware Grasp Re-Ranking for Task Oriented Grasping” 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
Manav Kulshrestha, Shreyans Jain, Apoorva Vashisth, Ankush Gupta, Damon Conover, Aniket Bera
LAGR separates physical grasp feasibility from task suitability by generating task-independent 5-DoF grasp proposals, then ranking the shared candidates using the instruction. The authors train and evaluate it with 2.68 million geometry- and rule-scored annotations. On the annotation-based benchmark, LAGR reaches 85.9% success versus 76.6% for GATER; this is not robot-execution success.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.