MINT-SJTUMIT

LA-33K

A language-action pretraining dataset for robot manipulation containing 33,116 atomic manipulation episodes derived from DROID and annotated with fine-grained language instructions. Designed to expose language-action supervision for training vision-language-action models.

Downloads47K
Likes1

Technical Profile

Modalities
rgbproprioceptionlanguage
Robot Embodiments
Franka Panda
Environment
lab
Task Types
manipulation
Data Format
LeRobot
License
MIT
Part of the LA-33K family

Community Signals

Top 5% by downloads

Access

Need custom rgb data?

Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets