This is the real-world SO-101 Bench demonstration dataset described in
SO-101 Bench: Measuring Object Generalization and Spatial Reasoning in
Vision-Language-Action Models. It contains teleoperated demonstrations
recorded with an SO-101 follower arm on a tabletop, using household objects
sourced from grocery stores, thrift stores, and the author's household.
The paper describes approximately 4,400 demonstrations (about 20 hours) over… See the full description on the dataset page:
https://huggingface.co/datasets/5hadytru/so101_bench_real_2.
