These are invented input examples, not a benchmark. No M expected-output/native parity assertion is made in this release. Validated historical examples are retained with XS and S.