Measured benchmarks and revised protocols
Release 0.5.0: a reproducible public-data result, better experimental design and a transparent publication status.
A real-data benchmark and corrected experiments
Release 0.5.0 · 20 October 2026. The original PanTHERIA archive yields 573 eligible mammal species. The fixed 10 kg hinge fails the primary held-out prediction screen; all sensitivities are retained. Acoustic and elastomer protocols are improved without inventing collected observations.
Explore the workstreams · Download the research archive · Download checksums
Visible changes
| Area | Change |
|---|---|
| Evidence | Measured species analysis, exclusions, family split, bootstrap, alternate error/split checks and two figures |
| Acoustics | Low-ka scope, independent tube allocation, uncertainty gates and design calculator |
| Elastomer | Correct softening sign, damping assumptions, acquisition requirements, thermal/probe controls and power sensitivity |
| Replication | Exact pulse recurrence checked against adaptive integration; no paper-figure replication claim |
| Publication | Current instructions, mixed-license attribution, author manifest and unsent review request |
Verify this deployment
The visible footer should say Release 0.5.0. Open version.txt and check build 2026-10-20-r5-measured. The deployment ZIP places index.php, assets, downloads and private directly at its root, with no public_html wrapper. Extract its contents into the domain document root.
This package was prepared and checked locally. It has not been uploaded to the live host by this update. After upload, verify the build and the measured benchmark page; clear any host/CDN page cache if the old build persists. Deployment and verification instructions.