experiment-forensics

Result verification validates reported research findings against the actual code and datasets used in the experiments. It applies cross-model code auditing and deterministic adjudication to ensure accuracy and reproducibility of third-party submissions.

You need to provide access to the experimental code, the reported results, and the necessary ledger file. Full verification requires the presence of both the evidence ledger and the experimental artifacts within the research paper's directory.

You audit a research paper's directory by running the experiment-forensics skill on it. The skill cross-checks the reported claims, code, and datasets against the evidence ledger to ensure the research is fully reproducible and valid.

Yes, experiment integrity auditing is designed to check third-party submissions for potential fabrications. It cross-references the submitted code, results, and ledger files to detect inconsistencies and validate the accuracy of reported findings.

If the ledger file is missing, the skill cannot perform full verification of the experimental results. Both the evidence ledger and the necessary experimental artifacts must be present to execute deterministic adjudication and detect discrepancies.