Caravan: Agent-Based Generative Data Labeling

Milestones

• Compatibility with OpenAI models
• Support for high-performance parallel data labeling
• Support for program-aided data labeling, i.e. rule cache in the Caravan paper
• Support for evaluation strategies like majority voting
• Application config file: Make sure that caravan-ai is application-independent

Reference

Please refer to our Caravan paper (published in OSDI 2024).

@inproceedings{zhang2024caravan,
  title={Caravan: Practical Online Learning of In-Network ML Models with Labeling Agents},
  author={Zhang, Qizheng and Imran, Ali and Bardhi, Enkeleda and Swamy, Tushar and Zhang, Nathan and Shahbaz, Muhammad and Olukotun, Kunle},
  booktitle={18th USENIX Symposium on Operating Systems Design and Implementation (OSDI 24)},
  pages={325--345},
  year={2024}
}

Contact

Qizheng Zhang ([email protected])