Policy
Data and code availability
What authors should share so that results can be checked and reused.
Expectation
Every article and book includes a data and code availability statement. Where possible, depositors place code, data and trained models in the Research Archive or another repository that issues persistent identifiers, and cite them in the reference list.
Reproducibility
Empirical work should report random seeds, number of runs, hardware, software versions, hyper-parameter search budgets and variance across runs. Reviewers in CONTINUUM Publishing assess reproducibility explicitly.
- State the exact datasets, versions and splits used.
- Report variability (standard deviation or confidence intervals) with the number of seeds.
- Describe compute used in accelerator-hours.
- Release evaluation code, or explain why it cannot be shared.