A quality-filtered, failure-labelled merge of SO-100 community manipulation datasets.
Built for imitation learning, RL from demonstrations, and failure-aware training.
The SO-100 open-source robot arm has generated dozens of small community datasets on HuggingFace, each contributed by independent researchers and hobbyists. Individually these datasets are too small and inconsistent to train on reliably.… See the full description on the dataset page:
https://huggingface.co/datasets/HaptalAI/so100-curated.
