About Potato
A lightweight, configuration-first annotation platform developed at the University of Michigan.
Our Mission
We believe annotation tools should be accessible to everyone, not just teams with dedicated engineering resources. Potato was designed to let researchers go from zero to annotating in minutes, using simple YAML configuration files.
Our goal is to democratize data annotation by providing a powerful, flexible, and easy-to-use platform that supports everything from simple classification tasks to complex multimedia annotation with AI assistance.
The Team
History
EMNLP Demo
Potato was featured as a system demonstration at EMNLP 2022.
Growing Community
Expanded features based on community feedback. Active learning support added.
AI Integration
Added LLM support with OpenAI, Claude, and Gemini integration.
Version 2.0
Major release with multimedia annotation, active learning, and enhanced UI.
Version 2.7
Potato began measuring how labels are made, not just what they are: an item-response-theory engine that gives every label a posterior and a confidence interval, live norming rooms with an agreement meter, counterfactual boundary probes, peer-prediction scoring, and local voice rationales — none of which need an LLM. The agent-evaluation suite grew to cover multi-agent teams and multimodal agents. Potato 2.0 was published at ACL 2026 (System Demonstrations).
EMNLP Demo
Potato was featured as a system demonstration at EMNLP 2022.
Growing Community
Expanded features based on community feedback. Active learning support added.
AI Integration
Added LLM support with OpenAI, Claude, and Gemini integration.
Version 2.0
Major release with multimedia annotation, active learning, and enhanced UI.
Version 2.7
Potato began measuring how labels are made, not just what they are: an item-response-theory engine that gives every label a posterior and a confidence interval, live norming rooms with an agreement meter, counterfactual boundary probes, peer-prediction scoring, and local voice rationales — none of which need an LLM. The agent-evaluation suite grew to cover multi-agent teams and multimodal agents. Potato 2.0 was published at ACL 2026 (System Demonstrations).
Citation
Cite Potato in Your Research
Official citation
@inproceedings{jurgens-etal-2026-potato,
title = "Potato 2.0: A Comprehensive Annotation Platform with {AI}-in-the-Loop Support",
author = "Jurgens, David and Chen, Michael and Iyer, Lina",
booktitle = "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
pages = "374--386",
url = "https://aclanthology.org/2026.acl-demo.37/",
}Related publications
Potato's design and capabilities are documented across these peer-reviewed papers:
- Potato 2.0: A Comprehensive Annotation Platform with AI-in-the-Loop SupportACL 2026 (Volume 3: System Demonstrations)
- Potato: The Portable Text Annotation ToolEMNLP 2022 (System Demonstrations)
Recognition: Best Demo Award, HCOMP 2024.
License
Potato is free, open-source software released under the GNU General Public License v3.0 (or later). You can use, modify, and redistribute it, including commercially, as long as derivative works stay under the GPL.
Have a question about deploying Potato or contributing back? Get in touch.
Contact the Team