petkopetkovother
Long-horizon subtask decomposition
A dataset for task decomposition in long-horizon tabletop manipulation, where models must produce ordered lists of pick-and-place subtasks from initial scene frames and compound instructions across SimplerEnv, RobotArena, and RoboLab benchmarks.
Downloads0
Episodes92
Why This Matters for Physical AI
This dataset enables evaluation of vision-language models' ability to decompose complex manipulation instructions into executable subtasks, a critical capability for autonomous robotic systems performing long-horizon goals.
Technical Profile
- Modalities
- rgblanguage
- Robot Embodiments
- WidowXFranka Panda
- Environment
- lab
- Task Types
- pick_and_placemanipulationtask_decomposition
- Episodes
- 92
- Annotation Types
- language_instructionsaction_labels
- License
- other
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack