---
license: mit
---
## Work-in-Progress
I trained the model with this data, and the L1 loss went down, but for some reason the model keeps failing the tasks.

## Citation
If you use this dataset or any part of the conversion pipeline, please cite the original RoboCerebra paper:
```bibtex
@article{han2025robocerebra,
  title={RoboCerebra: A Large-scale Benchmark for Long-horizon Robotic Manipulation Evaluation},
  author={Han, Songhao and Qiu, Boxiang and Liao, Yue and Huang, Siyuan and Gao, Chen and Yan, Shuicheng and Liu, Si},
  journal={arXiv preprint arXiv:2506.06677},
  year={2025}
}