Integrated GTCS Ablation Benchmark

This computational adjunct reports the first integrated ablation benchmark constructed around selected distinctions in the General Theory of Cognitive Structuring (GTCS). Its purpose is not to verify GTCS empirically or to convert formal distinctions into claims of universal computational necessity. Instead, a frozen synthetic realization is used to ask narrower questions: whether selected variables can be removed, collapsed, reparameterized, or behaviorally compiled away; whether observational confounding can make reduced models appear adequate; whether intervention-informed calibration identifies the intended causal role; how measurement noise changes recoverability; and whether high proxy-level predictive accuracy preserves a declared latent decision criterion outside calibration support. The benchmark produces mixed results. Current-state and finite recent-history rivals fail to reproduce history-sensitive outcomes in the frozen realization, supporting the computational utility of some history-bearing state information without establishing uniqueness of the GTCS retained-overload scalar. Pre-symbolic admission resists simple signalization and global-threshold collapses but is exactly reparameterizable by a context-dependent gate. An explicit transient coherence-representation variable is distinct in typed definedness but behaviorally compilable away for the single stateless downstream output used here. Observational models can achieve near-perfect in-distribution accuracy while failing at chance under frozen history interventions; recurrent recovery of a latent history state does not guarantee identification of its downstream causal role. Repeated measurement substantially improves out-of-distribution recovery, while a later criterion-transfer test shows that near-perfect calibration-support prediction can coexist with systematic misalignment from the intended latent criterion. The benchmark therefore supports a methodological separation among formal non-equivalence, computational necessity, state estimation, causal-role identification, support-level prediction, and criterion-transfer validity.