Repro Pack

Create a runnable, reviewable package, not just a README.

Baseline Reproduction

For reproducing a paper/method inside a research project, follow the Reproduction section of ai-research-workflow/references/project-workflow.md and its method template. Keep the target paper setting distinct from evaluation on the user's own dataset, inspect official code availability, fix data/evaluation versions, run a small smoke test, and verify the precommitted target before marking the result reproduced. Preserve partial/failed/stopped runs and deviations. Formal run reports use experiment-report-writing; data source/copy handling uses research-dev-standards. Package artifacts for handoff/release when that is the requested deliverable.

Reproducibility Package

Recommended structure: