H4 ablation run records and specifications: controlled tests of Interface, Execution Pathway and Authority correspondence
Description
Run records, fixed experiment specifications and execution logs from fourteen controlled ablation configurations testing whether relational structures classified under Interface, Execution Pathway or Authority admit the predicted ablation signature while the other two conditions are independently verified unchanged, together with eight Cut tests of whether removing a supplying node denies a condition exactly where the node lies on every available supply. Each configuration fixes, before its baseline is established, the operation and its transition, the nominated interface, the acting mechanism and its execution identity, the node and governance partitions, the resolution depth, the single permitted ablation and the predicted post-ablation values. Configurations coded at both a system and a system-of-systems attribution are executed separately under each coding; these are the same technical mechanism under different governance attributions and are not independent replications.
Files
Steps to reproduce
manifests/ holds the fixed specifications and predictions. experiments/<ID>/results/<coding>/run-NN/ holds per-run evidence: the run record, the resolved container composition, and the baseline, post-ablation and access-state observations. logs/ holds execution logs including dry runs. demonstration-cases/ holds the per-case assessment records for the five incident and technique analyses in the associated paper. The Docker and Python harness that produced the records is included. Results directories contain more runs than the reported batch of five per coding. The additional executions are calibration runs, or runs in which the execution or evidence-collection scaffolding failed to satisfy the specified evidence rules; these are retained deliberately and excluded from the reported batch. A run is technically invalid where the scaffolding failed to produce evidence meeting the rules, never where a prediction was unmet.