spikefly/SemanticVLA-BCZ-LeRobot

SemanticVLA · BC-Z (LeRobot + dense traces)

🚧 Placeholder. The URL is stable; data will be uploaded incrementally per the release roadmap.

LeRobot conversion of the BC-Z dataset (Open X-Embodiment slug bc_z), paired with dense end-effector trace annotations. Used to train the trace head and the trace-conditioned Latent Action Model (LAM) in SemanticVLA.

Field	Value
Robot	Everyday Robots manipulator (same family as RT-1)
Episodes	39,350
Frames	5,471,693
Cameras	front RGB (224×224 in LeRobot conversion, 5 fps H264)
Action space	future/xyz_residual(3) + future/axis_angle_residual(3) + future/target_close(1)
Trace alignment	per (episode, step), present for 100% of episodes

Related repositories

Resource	Link
Code	https://github.com/Fei-Ni/SemanticVLA_Offcial
All-in-one dataset (4 subsets bundled)	https://huggingface.co/datasets/spikefly/SemanticVLA-TraceX-260K
Sibling dataset · Bridge	https://huggingface.co/datasets/spikefly/SemanticVLA-Bridge-LeRobot
Sibling dataset · Fractal	https://huggingface.co/datasets/spikefly/SemanticVLA-Fractal-LeRobot
Sibling dataset · DROID	https://huggingface.co/datasets/spikefly/SemanticVLA-DROID-LeRobot
Checkpoints	https://hf.co/collections/spikefly/semanticvla-model-zoo
Collection · Datasets	https://hf.co/collections/spikefly/semanticvla-datasets
Collection · Model Zoo	https://hf.co/collections/spikefly/semanticvla-model-zoo

Planned layout

SemanticVLA-BCZ-LeRobot/
├── meta/                       # info.json, episodes.jsonl, tasks.jsonl, stats.json, modality.json
├── data/                       # per-episode parquet shards (39,350 episodes)
├── videos/                     # per-episode H264 mp4s (39,350 episodes)
└── traces/
    ├── bcz_coords.npy
    ├── bcz_offsets.npy
    └── bcz_present.npy

How to use (placeholder API)

from huggingface_hub import snapshot_download

local = snapshot_download(
    repo_id="spikefly/SemanticVLA-BCZ-LeRobot",
    repo_type="dataset",
)

Citation

@misc{semanticvla2026,
  title  = {SemanticVLA: Semantic Vision-Language-Action Models with Trace-Conditioned Latent Action Tokens},
  author = {SemanticVLA Authors},
  year   = {2026},
  url    = {https://github.com/Fei-Ni/SemanticVLA_Offcial}
}

@inproceedings{jang2021bcz,
  title     = {BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning},
  author    = {Jang, Eric and Irpan, Alex and others},
  booktitle = {Conference on Robot Learning (CoRL)},
  year      = {2021}
}

License

Released under the MIT License, subject to the upstream BC-Z / Open X-Embodiment licenses.