Prior art first
Research starts by trying to show that the contribution already exists in a strongest-reasonable composition of prior art. An apparently unoccupied space does not prove value or novelty.
Real objects
Real or representative objects force architecture to respond to evidence not designed to favor it. Conditions and limits should be frozen before comparison.
Negative controls
Negative controls help detect whether a method produces favorable signals even when it should fail. A useful evaluation system must show that it protects something real.
Null results
If the incumbent solves the case equally or better, or evidence does not distinguish a material contribution, the null result is preserved. The case is not rewritten until the architecture “wins.”
Responsible publication
Public communication clearly separates proposal, implementation, verification, standard and scientific contribution. Publishing less but supporting it better is part of the discipline.