Generated Prediction and Interval Artifacts for the C-MAPSS RUL Evaluation Audit

Generated model predictions, model checkpoints, run metadata, and uncertainty-interval artifacts underlying the controlled C-MAPSS remaining-useful-life evaluation audit. The deposit contains 120 registered Kill Test runs, 160 unified point/quantile runs, and 280 post-processing uncertainty cells, together with aggregate result tables and validation schemas. It also carries the aggregate outputs and frozen plan for secondary common-truth rescoring of the unchanged Kill Test predictions. The pre-listed reversal check was LSTM versus CNN; any additional model-pair reversal from the exhaustive implementation scan is exploratory. Checkpoints are retained so the archived run contract remains complete; the reported post-hoc analyses consume the prediction artifacts directly. This is derived research output. It does not contain or redistribute NASA C-MAPSS source sensor tables, official source archives, literature PDFs, personal data, or credentials. Prediction tables retain NASA-derived unit/cycle identifiers and target values needed for validation. CC BY 4.0 applies to the creator's copyrightable contributions; no ownership or new license is asserted over NASA-origin facts, which remain subject to applicable upstream terms. Use the companion software record to validate and analyse the files.