Use when strengthening reproducibility for ICLR papers, including seeds, variance, compute, datasets, implementation details, ethics statements, and reviewer-verifiable evidence. Use when writing the ICLR reproducibility statement, when a reviewer says a result is not verifiable, or when mapping each representation-learning claim to a seed, split, and command so anyone reading the permanent OpenReview record can check it.
