Research record
Current evidence, protocols, and claim boundaries
This page points to the current public research record rather than treating older papers as automatically authoritative. Start with the claim boundary, then follow the promoted evidence packages, preregistered protocols, and reproducibility records for each result.
Updated July 2026
Current Claim Boundary
The current supported claims, withdrawn interpretations, active evidence families, and required next evidence for the research program.
Start here
July 26, 2026
Corrected Paired-Acquisition Evidence
The forward-valid corrected release for fold-aware SCORPION inference and the corrected fixed-estimand canine SCC audit.
Corrected release
July 26, 2026
Capacity-Matched SCORPION Ablations
A seven-variant, five-fold, five-seed campaign using a true equal-capacity two-branch control and registered objective ablations.
175 / 175 fits
July 27, 2026
Dimensionality × Cross-Covariance Factorial
A preregistered canine SCC factorial that found no universal dimensionality effect or stable operating point under the tested protocol.
450 / 450 cells
2026
Pair-Repeat Allocation Effects
A matched-budget study of whether paired-acquisition learning benefits more from biological pair diversity or repeated anchor exposure.
6,400 / 12,800 budgets
July 29, 2026
Prospective Paired Affine Comparison
A leakage-resistant comparison of neural factorization with centroid translation, orthogonal Procrustes, affine least squares, and ridge affine controls. No result is claimed yet.
Execution pending
Open research record
The repository contains the code, frozen protocols, result artifacts, provenance records, and explicit boundaries needed to audit the active claims.
Computational pathology research repository
Main codebase, current README, evidence packages, technical documentation, and reproducibility infrastructure.
Research-engineering brief
A concise explanation of Paired-Acquisition Neural Factorization, its design, current evidence, and limitations.
Scientific audit remediation ledger
The corrections, withdrawals, reruns, and evidence-promotion status behind the current public record.
Study package index
Focused SCORPION, external canine SCC, and pair-allocation repositories and documents.
Claim boundary. This is research-only work under controlled experimental conditions. It is not clinically validated, diagnostic software, or intended for patient-care decisions. The public paper remains on scientific-audit hold; the repository claim boundary and promoted evidence packages are authoritative when older summaries conflict.