Gravitational Manipulation Theory v2.2 — Information-Ablation and the Minimum Record for Constraint Reconstruction

Gravitational Manipulation Theory (GMT) v2.2 asks a constructive question: what is the minimum set of published information sufficient for an external reconstruction of a precision-force constraint to within a fixed tolerance? A fully known Controlled Synthetic Apparatus (CSA) is introduced as ground truth. An information-ablation experiment is then performed over a nested hierarchy I0 ⊂ I1 ⊂ I2 ⊂ I3 ⊂ I4, progressively removing component geometry, absolute geometry, response-operator information, and statistical-mapping information. Operator ambiguity D_H and constraint ambiguity D_C are quantified as log-space (dex) spreads over apparatus configurations consistent with the disclosed information. The tolerance ε_H = ε_C = 0.30 dex, priors, sampling protocol, and certification rule were frozen before the ablation. The measured constraint-ambiguity curve is D_C = (0.000, 0.511, 0.738, 5.007, 6.069) dex for I4 → I0. The minimum sufficient information level is I_min = I4, while operator sufficiency extends down to L*_operator = I2. The resulting CSA verdict is Ceiling Conditionally Closed. Every PASS is supported by an exact analytic range; Monte-Carlo estimates are treated as lower bounds and are used only to certify FAIL when they already exceed the pre-registered tolerance. The sufficient layers are additionally translated into a practical publication-facing minimum-record checklist. This checklist is an expansion of the measured information layers, not a claim that each individual field was separately ablated. The resulting minimum is class-conditional and tolerance-dependent; the information-ablation methodology, rather than the numerical minimum itself, is the portable result. Finally, the HUST-2020 public record is evaluated as an external-record audit against the same minimum-record rubric. Several apparatus-native fields are not recoverable from the public record examined, yielding an audit verdict of Ceiling Retained for that record. This is strictly an audit of recoverability from the specified public record: HUST is never treated as ground truth, unpublished apparatus values are never inferred as true values, and the published physical result is not re-evaluated. The CSA is synthetic and representative-class, not HUST hardware. The representative Yukawa-form force kernel is used only to generate the controlled synthetic response and is not introduced as a new physical constraint. No experimentally demonstrated gravitational manipulation is claimed in this work. The supplementary archive contains the CSA and ablation code, machine-readable validation and ablation results, frozen pre-registration materials, Stage 1 and Stage 2 reports, and Figures 1–7.