Frozen predictions and an evidence-led review
Release 0.7.0 improves the proposed elastomer test, exposes existing model weights and keeps external completion claims tied to records.
Prediction first
Release 0.7.0 · 20 October 2026. The elastomer protocol now makes frozen prediction on new independent specimens its primary target. A conventional softening reference, scoring code, uncertainty method and review-resolution record make the proposal easier to inspect.
Predictive protocol · Review decisions · Research-only archive · Download checksums
What changed and what did not
| Area | Release status |
|---|---|
| Predictive design | Independent-unit loss, frozen rivals, relative/absolute criteria, joint calibration uncertainty and explicit unresolved freeze fields. |
| Code | Twelve implementation checks and three synthetic decision controls; no material fit. |
| Reporting | Measured AICc weights shown; 573-species primary result unchanged. |
| Sources | 134 records; new official softening documentation. Earlier reading-depth limits remain. |
| Publication | Prepared archive; no new external registration, DOI, review or submission performed. |
Verify the deployed version
After extracting the flat ZIP into the domain document root, open version.txt with the new build marker. It should say Release: 0.7.0 and Build: 2026-10-20-r7-predictions. The footer also shows 0.7.0.
This package is prepared locally, not uploaded to the live host. Deployment and cache checks explain how to distinguish an old ZIP, an incorrect extraction folder and cached pages.