A convincing-looking Unreal validator was rejected because it reconstructed transforms from the source CSV instead of observing instantiated geometry. Three planted geometry corruptions left its result byte-identical.

The system said: The transforms invert to the source values; validation passed.
The evidence showed: Both sides came from the source row. Three changes to instantiated geometry were invisible to the validator and visible to an independent readback witness.

§ 1

The record

FINDING · unreal-validator-never-read-instantiated-geometryobserve · observation boundary

A convincing-looking Unreal validator was rejected because it reconstructed transforms from the source CSV instead of observing instantiated geometry. Three planted geometry corruptions left its result byte-identical.

Claim: The D495 V0 validation path could not establish correspondence between source values and instantiated Unreal geometry: its transform check calculated both the placed value and the inverse from the same CSV-derived row, and its runtime relocation path continued to iterate source rows without geometry readback. In a 4,225-sample regression fixture, three post-ingestion geometry corruptions changed the independent witness by up to 4.5 metres, 44 metres and 5 metres while the D495-style check continued to report zero error and produced byte-identical output.

Status: OBSERVED Witnessed in one instance. No claim about how often.

Question: Passed is not correct

Subject: An internal D495 Unreal validation workflow and a local regression fixture; the qualifier, planted corruptions and independent readback witness are Omega's. built for the study

Frame: One 4,225-sample fixture reconstructed from the D495 dependency pattern; one clean control and three post-ingestion corruptions: scale x1.1, a 400-vertex zeroed patch and a +5 metre shift; seven authored instrument-dependency cases in the qualifier prototype.

Method: Trace each claimed comparison to the source of both operands, then mutate the claimed subject after ingestion while keeping the source unchanged. Compare the original validator output with a witness that reads the instantiated geometry rather than reconstructing it.

Oracle: Direct readback of the mutated instantiated coordinates in the regression fixture, compared with the frozen source coordinates. The original validator's own transform and inverse are not used as the oracle.

Negative control: Present Three planted geometry corruptions had to be detected by the readback witness and missed by the source-derived check; the clean control had to remain at zero. The witness reported maximum errors of 4.5 metres, 44 metres and 5 metres, while the invalid check remained at zero and byte-identical.

Denominator: One clean control and three planted corruptions over one 4,225-sample fixture; seven authored dependency-graph cases in Instrument Qualifier v0.

Preregistration: None. This was an instrument audit and local regression investigation, not a preregistered experiment.

Limitation: This does not show that the original Unreal scene was geometrically wrong, and no repaired Unreal validator was built or run against the scene. The regression fixture reproduces the data dependency locally; it is not an independent reproduction in Unreal. Instrument Qualifier v0 used seven authored dependency graphs, so its recall on real validators is unknown and no general qualification capability is claimed.

Source: artifact not public

Reproduce: not reproducible from a public clone

Independent reproduction: None known.

§ 2

Where this sits

This finding answers Passed is not correct and supports the OBSERVE stage of the operating method. It is an instance of the observation boundary mechanism.

Evidence ledger Bring us a problem