Preregistration (Version 1.0) for a computational evaluation study testing whether an HDT²/GIT runtime inquiry pipeline improves domain-bounded AI reasoning beyond base-model use, retrieval augmentation, standard structured prompting, and generic multi-stage scaffolding. The study uses ablation-based primary analysis across five preregistered behavioral metrics with locked effect-size thresholds, neutral-label controls, and predefined pass/fail conditions. Methodological inspiration drawn from Vargas-Barroso et al. (2026, Nature Communications) without claiming biological equivalence. No amendments filed as of deposit.
Bruce Tisler (2026) studied this question.