“…Even though many automatic software systems for handling large proof objects are being actively developed, additional work is needed to identify and remove possible redundancies, so that the extracted programs are more readable, shorter and faster. Examples of such optimisations include uniform decorations [1,7,12], soundness-preserving program transformations [10], avoiding syntactic repetition [16]. In the present paper we suggest another such technique for the Dialectica interpretation, which marks computed counterexamples that are determined to be valid.…”