CUT test: exclusivity measurement
Candidate Uniqueness Test: measures whether the rule set forces one dominant reading per EVA token or admits several equally good ones. Phase 1 scores candidates against the lexicon (large TopGap = unambiguous), Phase 2 generates readings from the formalized rules. The figures measure the internal unambiguity of the rule set, not whether the text is Hebrew: meaningless control texts so far reach the same values as the manuscript (see model comparison).
Critical ≤ 15Warning zone 15–20Target > 20 (N ≥ 300)