Open Models Almanac
Methodsdata curationdata sourcing

specific method · filed under data curation

Pre-training on community-contributed robotics datasets

Pre-trains on publicly available, community-contributed robotics datasets, with evidence describing curated mixtures or use of such datasets but not a single standardized recipe.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See data sourcing for the methods filed beside it.

Also called community-collected datasets, Pretraining on community robotics datasets, pretraining on community datasets.

sources
2
model
1
lab adopt it
1
strongest
core

Documented in

Hugging Face

Evidence

3 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
evaluated 1used 1core 1

SmolVLA

Hugging Faceadopted2 sources

It is pretrained on a curated mix of publicly available, community-contributed datasets designed to reflect real-world variation.

coredata curationin SmolVLAHugging Face
2 more from this model

SmolVLA is trained end-to-end on fewer than 30k episodes drawn entirely from publicly available, community-contributed datasets

useddata curationin SmolVLAHugging Face

pretraining on community datasets leads to a substantial performance improvement

evaluatedunclearin SmolVLA
↑