AIPOCH-AI/open-science-cases

Open Science Cases

Private case library for AIPOCH product development, evaluation, and reproducible demonstrations. Store each case's task, source inputs, expected results, and observed outputs together.

This is the dataset repository: datasets/AIPOCH-AI/open-science-cases. Use this destination for case submissions.

Submit a case

1. Name the case folder after its case title, using lowercase words joined by hyphens (for example, smoking-and-depressive-symptoms-in-nhanes-2015-2016). Check that the name does not already exist; add a short suffix if needed.
2. Create cases/<case-name>/. The worked example is in cases/template/.
3. Copy SUBMISSION_TEMPLATE.md to that folder as README.md. Replace placeholders; use N/A where a section does not apply.
4. If the case comes from Open-Science, export its session as session.science and add PROMPT.md with the original user request and meaningful user-authored follow-ups. Keep the full research history in the package.
5. Add inputs and supporting files following DIRECTORY_STRUCTURE.md. Create only the directories needed for that case.
6. Upload the whole case folder together. Use a commit message such as Add case: <case-name>.
7. Start with status draft. When complete, set ready-for-review; a reviewer records their findings and marks it accepted or needs-work.

These statuses are documentation conventions, not automated approval gates. Keep the same folder name when revising a case; Git commits record its history.

Upload from the CLI

After logging in with a token that has write access to this organization:

hf upload AIPOCH-AI/open-science-cases ./cases/<case-name> cases/<case-name> --repo-type dataset

Replace <case-name> before running. For browser submissions, use Files → Add file → Upload files, preserving the case directory paths.

What belongs here

• Product scenarios, prompts, source documents, reproducible steps, and expected outcomes.
• Actual outputs and evaluation notes tied to a product/model version.
• Supporting images, PDFs, tables, or other files referenced by a case.

Record source URLs and usage restrictions for third-party materials. Repository privacy does not grant rights to redistribute source material. Repository content is licensed under the Apache License 2.0, except where third-party materials have their own licenses or usage restrictions.

Reading cases

Browse the cases/ directory and open a case's README.md. This is a folder-based case archive, not yet a standardized table for load_dataset() or the Dataset Viewer. For Open-Science cases, open PROMPT.md for a concise request history and import session.science into a compatible version of Open-Science to inspect the complete package.