For technical evaluators
This page is for the people asked to check us: CTOs, security reviewers, engineers and scientists. It describes the method at the level of what you see and do, not how it works inside.
Method
What we are building toward, and where you decide
- How results are presented
- The intended result separates what was observed from what is inferred. Observed data is drawn solid; inferred data is drawn dashed. The aim is to keep the data behind a result available for review.
- How confidence is shown
- Our intended results carry one of three plain labels: High confidence, Worth checking, or Not enough information. When information is missing, the system should ask instead of guessing.
- Where people decide
- The intended system recommends. A named person will accept, reject or change each result, with that decision recorded alongside it.
- How corrections are handled
- We are working toward learning from an expert's correction without a retraining cycle, with the correction and who made it kept on record. PALM is in internal testing; PIE and VILLA are not yet feature-complete.
Under NDA
What we share under NDA
- A live walkthrough with a member of our research team
- Detailed documents for security and data-protection review
- [Other material Hyuge shares under NDA]
To request them, tell us who you are and what you're evaluating. We'll send an NDA, and once it's signed we'll share the material.