Human Feedback Loop¶
This page has moved
The feedback loop is now documented in full — flag, curator gate, episodic memory, re-train and promotion — on Self-Evolving Security.
Geodesia turns everyday usage into a continuous improvement flywheel: a user or reviewer flags a bad decision, a curator approves it, and the correction lands on the timescale that matches it — a threshold, an episodic memory consulted at scoring time, or the model weights.
| I want to… | Go to… |
|---|---|
| Understand the whole loop and where each part runs | Self-Evolving Security |
| Raise a flag / review one / export the corpus (API) | Self-Evolving Security → REST API |
| Turn on the episodic memory | Self-Evolving Security → the fast loop |
| Trigger a re-train and promote a checkpoint | Self-Evolving Security → the slow loop |
| Tune a threshold on real traffic instead of correcting one case | Policy Lens |