Finite anti-checkers #
This module exposes finite anti-checkers as concrete lists of inputs. A list
anti-checks a target at threshold s when every circuit of size at most s
disagrees with the target on a listed input. The predicate is monotone under
adding inputs, antitone in the circuit threshold, and invariant under list
permutation, so the list acts semantically as a multiset.
The main bridge is exact: the canonical SuccinctMCSP instance labelled by the target rejects if and only if its input list is an anti-checker. The extraction layer obtains such a list from any finite covering set of circuit codes whose members all fail somewhere, using at most one disagreement input per code. The enumeration layer supplies a canonical covering set with an explicit exhaustive cardinality bound. Its survivor-count layer identifies anti-checking exactly with reducing the number of consistent canonical candidates to zero. Relative estimates of that count use an exact cross-multiplied natural-number contract, and minimizing those estimates preserves survivor shrinkage with explicit loss. Shrink traces compose those one-round bounds exactly. The good-string layer supplies exact tuple counts and finite averaging.
Adding possible counterexample inputs preserves anti-checking.
An anti-checker for a larger circuit class also anti-checks every smaller threshold.
Padding with all-zero inputs preserves the anti-checker property.
A circuit satisfies every canonical sample induced by an input list exactly when it agrees with the target throughout that list.
A list is an anti-checker exactly when no circuit within the threshold matches all samples in the corresponding typed SuccinctMCSP instance.
Exact encoded rejection bridge between anti-checkers and SuccinctMCSP.