Research Question
Persistent derived state can retain apparently admissible support after an upstream dependency has been omitted. VCT operationalizes separate obligations for support enumeration and lineage-output coverage, then examines how release policies behave when clean alternative support may also exist.
The project tests whether a specific omitted-lineage failure can be made executable, measured, and falsifiable within declared boundaries.
Why This Matters
Persistent derived state can retain apparently admissible support after an upstream dependency is omitted. An executable case makes this failure inspectable and separates continuity integrity from authorization control.
Experimental Approach
The internal suite contains 520 structured coordinates derived from 40 truth worlds. Clean alternative support is deliberately present in designated strata, and individual claim evaluations are not independent real-world observations.
VCT compares its V2 policy with two conservative controls: one defers uncertain claim scope, while the other requires every candidate lineage to be complete. Neither control represents every plausible simpler alternative-aware policy.
Current Evidence
Within the frozen 360-coordinate truthful-incompleteness subset, V2 released 10,440 of 12,960 evaluator-valid claims. Each tested conservative control released 8,160. The difference occurred only where clean alternative support was available.
Current Claim
Within the frozen truthful-incompleteness subset, V2 released more evaluator-valid claims than the two tested conservative controls, with the same observed invalid-release count. The difference occurred only where clean alternative support was available.
The external reduced-filter pilot did not show a reduction in unauthorized requested submissions.
Claim Boundary
VCT produced a reproducible testbed for asking a bounded question: can a continuity system detect controlled corruption, recover the affected lineage, and preserve legitimate changes made afterward?
On the frozen synthetic suite, VCT operationalized application-specific lineage-coverage requirements and compared three recovery policies. It also preserved executable counterexamples that can expose unsupported claims of successful recovery.
The central result was negative: lineage closure and local containment did not reduce unauthorized actions under the tested external-action measure. Knowing where corrupted information came from—and containing or repairing that portion of the continuity record—was not enough to keep subsequent behavior within the intended authority boundary.
What that means
VCT provides evidence that continuity integrity and agency control should be treated as separate engineering problems. Provenance and recovery mechanisms may improve the integrity of memory state, but they should not be treated as substitutes for explicit authorization enforcement.
The preserved contribution includes the evaluation instrument: a reproducible way to test recovery claims and retain failures, counterexamples, and legitimate post-checkpoint changes.
Limits
VCT does not establish conceptual priority for provenance completeness, general portability, the external effectiveness of the complete VCT V2 architecture, superiority over every simpler policy, or a general solution to agent-memory safety.
Active academic-publication development stopped after adversarial review. The artifact was preserved because the bounded case, internal comparison, negative result, counterexamples, and stopped repair path remain technically useful.
Negative or Null Results
When coverage was falsely marked complete, every tested policy released 455 of 1,440 evaluator-invalid claims. A separate EAL-Bench Procurement pilot then tested a reduced exact-identifier citation-coverage filter—not full VCT V2.
| Condition | Unauthorized requested submissions | Authorized use |
|---|---|---|
| Baseline | 11 / 44 | 36 / 36 |
| Reduced filter | 11 / 44 | 35 / 36 |
The pilot showed no observed reduction in unauthorized requested submissions. Pairing limitations prevent clean causal attribution of the single authorized-use difference, and the primary safety metric excluded an additional unauthorized alternative action by the filtered arm.
Next Falsification / Validation Step
Active academic-publication development stopped. No additional experiment is justified merely to increase publication appeal. A future external semantic-support experiment would require its own research question and validation chain.
Artifacts and Publications
The technical note presents the working claim, methods, quantitative boundaries, external pilot, and mandatory nonclaims in proportion to the evidence.
Open technical noteReview History
The published record states that active academic-publication development stopped after adversarial review. It does not provide a dated review log here; no review dates or verdict identifiers are added.