FineART contains 1,718 hours of two-handed robot demonstrations divided into 533,913 labeled subtasks.
Dense subtask labels help robots complete long two-handed tasks
Training on 534,000 labeled subtasks improved spatial instruction following, long-horizon execution and transfer to new robot hardware.
AI Startup
Jade Choghari · Pepijn Kooijmans · Mansi Agarwal · Yusuf Umut Ciftci · Aseem Doriwala · Catherine Weaver · +5 more
Scale AI · Hugging Face · University of Southern California · Stanford University
Research Digest··2 min read
The authors assembled FineART, a bimanual manipulation dataset containing 40,543 episodes across 151 tasks, with language annotations marking more than half a million intermediate steps.
Why this paper
From Hugging Face and 3 others
In one line
FineART, a bimanual manipulation dataset with dense subtask annotations, enables a vision-language-action model to achieve 100% success on spatial instructions.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (5 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§