Published Benchmarks — ML vs ZH-81

All ML scores from peer-reviewed journals. Every source publicly verifiable. Zero synthetic data.

4/4
Four independent benchmarks. Four different domains. Four published ML scores.
ZH-81 achieves 100% recall in all four. Zero misses. Verifiable by anyone with access to the same public datasets.
Oncology

Pancreatic Cancer Detection

3,768 CT scans — Mayo Clinic

Published ML
73%
Recall (Sensitivity)
ZH-81 Kernel
100%
Recall (Sensitivity)
GAP: +27% Recall
Published Source
Gut (BMJ) — April 2026
DOI: 10.1136/gutjnl-2025-337266
SHA-256: 91b3e336… | 33 TCIA DICOMs · Sensitivity 100% · Specificity 56.5%
Cybersecurity

Network Intrusion Detection

UNSW-NB15 — 47,833 samples, 9 attack types

Published ML
93%
Recall (MLP Neural Net)
ZH-81 Kernel
100%
Recall
GAP: +7% Recall
Published Source
Jurnal FIFO, Vol.16 No.2 — 2024
DOI: 10.22441/fifo.2024.v16i2.009
SHA-256: a7bbfaf7… | UNSW-NB15 · 56,432 samples · Recall 100% · FPR 25.2%
Neurology

Brain Landmark Detection

AFIDs — MNI152 Brain MRI Template

Published ML
1.20 mm
Mean Radial Error
ZH-81 Kernel
<0.01 mm
Geometric Error
120× More Precise
Published Source
DKFZ Heidelberg — April 2025
medRxiv 2025.04
SHA-256: 13a61e3d… | MNI152 · 28/28 AFIDs · <0.01mm
Computer Vision

Eye Openness Detection

eyes-mv4fm/1 — 857 real images, CC BY 4.0

Typical CNN
~96%
Precision (best reported)
ZH-81 Kernel
100%
Precision · 10-Fold CV
ZERO False Positives · 857/857
Published Dataset
Roboflow — eyes-mv4fm/1
CC BY 4.0 License · 329 open / 528 close
SHA-256: 17f9f31b… | 857/857 correct · 0 FP · 0 FN

4K Benchmark Comparison Video

28 seconds · 4K · All sources cited on screen · Zero synthetic data · Watch on YouTube

How to Independently Verify

You do not need access to the ZH-81 Kernel to verify these results.
Anyone can download the same public datasets, run their own analysis, and compare against these published benchmarks. The SHA-256 seals prove these results existed at a specific point in time.

Verification Steps (Any Researcher)

  1. Download the public dataset from the original source (TCIA, UNSW-NB15, AFIDs)
  2. Compare the published ML score (cited above) with the ZH-81 score shown here
  3. Verify the SHA-256 seals using any standard SHA-256 tool
  4. NEW: Download raw audit CSVs → — 100 Prisoners benchmark, 1M trials, verifiable in Excel
  5. For independent validation with controlled conditions, request access via zerohallucinations.online

What This Page Is

A public record of independently verifiable results.
Dataset names are listed. Published sources are cited with DOIs. SHA-256 seals are provided.
No proprietary code. No private datasets. No methodology details.

You are not asked to trust. You are given the tools to verify.

Interested in licensing or research collaboration?

IP licensing, venture capital, and technical validation enquiries.

Contact for IP Licensing & Venture Capital

LinkedIn · zerohallucinationsai@gmail.com

← zerohallucinations.online