Anonymization Notes for Double-Blind Review
Scope
This repository has been prepared for anonymous reviewer access without changing benchmark logic, generator logic, model logic, labels, matched-prefix evaluation, or reported results.
What Was Anonymized
- Local absolute filesystem paths were removed from review-facing documentation and replaced with repository-relative references or
<repo-root>. - Review-facing metadata notes were updated to avoid personal machine paths.
- Anonymous placeholders are used where author or institution details may need to be restored later:
Anonymous AuthorsAnonymous Institution
How To Reproduce Results Anonymously
- Clone or unpack the repository into any local directory.
- Install dependencies from
requirements.txtorenvironment.yml. - Run experiments from the repository root using relative paths only.
- Use the deterministic settings documented in
docs/DETERMINISM.md. - Use the released benchmark configurations and paper-suite result files without editing benchmark code.
- For double-blind sharing, distribute a source archive of the working tree rather than the
.git/directory, since git history and config may contain identifying metadata.
What Will Be De-Anonymized After Acceptance
- Author names:
Anonymous Authors - Institution names:
Anonymous Institution - Public dataset repository:
https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins - Public code repository:
https://huggingface.co/temporal-twins-benchmark/temporal-twins-code - Croissant metadata:
https://huggingface.co/datasets/temporal-twins-benchmark/temporal-twins/raw/main/metadata/temporal_twins_croissant.json - Paper or preprint: not available during double-blind review and to be added after publication
- Any optional acknowledgments withheld for double-blind compliance
Data Statement
Temporal Twins contains no real UPI data, no real users, no real bank accounts, no real transactions, and no personal financial records. The benchmark is fully synthetic and is intended only for research evaluation.