Errata and supplement to Technical Report v1. 10. This document corrects the empirically-detected texture-axis transition point σₛtat reported in Technical Report v1. 10 for Upscayl Standard, computed against a database later found to contain a data-integrity bug in its population-filtering logic: a third, more severe variant of a previously-documented join-miss artifact in this database schema, in which a comp_* feature-extraction pass attached a different, non-zero block's values to the wrong row — producing plausible-looking but incorrect numbers that the existing iscompᵦero flag cannot detect, since the resulting value is genuinely non-zero. An independent, from-scratch verification database (no join step at all — every value computed directly from the source images in a single pass) was built for both Upscayl Standard and SwinIR and used to re-derive σₛtat under the exact original methodology, isolating the filter as the sole variable changed. Result: Upscayl Standard's σₛtat is corrected from 0. 0800 to 0. 101 (+0. 021) ; SwinIR's σₛtat is confirmed unaffected at 0. 0800. As a due-diligence cross-check, the luminance-axis statistic μₛtat — which depends on the same iscompᵦero flag via a different (inclusion, not exclusion) usage — was also re-verified for both models and found unaffected (Standard 0. 0500, SwinIR 0. 0100, unchanged). As an additional cross-check on the new database's own internal correctness, blocks where SSIM (GT, NOTTA) =1. 0 and SSIM (GT, TTA) =1. 0 hold simultaneously (a case where, by a Pixel Identity theorem, both reconstructions are forced to be pixel-identical to the ground truth, and therefore to each other) were confirmed to match exactly on all eight comp_* features between GT, NOTTA, and TTA, with zero exceptions across 702, 108 such blocks (Standard). Both models were verified via live SQL execution against the new databases, screen-recorded (one continuous recording per model) and hash-verified. The complete database schema, verification database build pipeline, and verification SQL are included to enable independent reproduction. Upscayl High-Fidelity and Digital-Art are outside this report's scope. Dataset and Mandatory Citation: - Source: NIH ChestX-ray8 (Hospital-scale chest x-ray database) - Citation: Wang, X. , Peng, Y. , Lu, L. , Lu, Z. , Bagheri, M. , & Summers, R. M. (2017). "ChestX-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thoracic diseases. " Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 3462–3471. - Download: https: //nihcc. app. box. com/v/ChestXray-NIHCC Contact: s. shiny. n. works@gmail. com
neco mohumohu (2026) studied this question.