The dataset behind robot foundation models.
A separate product from the Loop platform: large-scale converted human action data for pre-training, paired with high-precision robot teleop data for post-training — drawn from Config's pipeline, bought outright.
130K+
Total hours · human + robot
8K+
Tasks across environments
7M+
Episodes ready to train
Pre-train at scale. Post-train with precision.
Robot foundation models today need two kinds of data — and Catalog gives you both. Pre-train on hundreds of thousands of hours of human action data, captured at industrial scale through our Hanoi Data Studio and converted to your target embodiment through Config’s Conversion Engine.
Then post-train on high-precision robot teleop data from our Seoul lab — already in robot embodiment, ready to fine-tune on directly. Browse by source, embodiment, and task. Delivered in LeRobot format.
- humanUMI · 2F-85cloth folding1,240 h
- robotFR3 · 2F-85pick & place — recyclables412 h
- humanUMI · 2F-85desk organization890 h
- robotVega · Vega Handcleanroom assembly186 h
- humanBare handfood preparation620 h
- robotRB-Y1 · 2F-85coffee serving94 h
SAMPLE · 8K+ TASKS AVAILABLE
Talk to our team.
Tell us about your task, your robot, or the data you need. We’ll show you how Config fits.