# RH AI Research Starting Point v2.2

## Node

`RH-W-19-REPRODUCIBILITY-AND-ADVERSARIAL-AUDIT`

## New Additions

- Unified certificate backend v0.2;
- 16 classes of rejectable false certificates;
- 1 class of verifier collusion attack expected to survive;
- exact Hilbert-14 floating-point false negative;
- External signatures, reproducible builds, and independent verifier roadmap.

## Progress

Batch 01 has completed `RH-W-19/20`.

`RH_CLAIM=False`.