Skip to content

Jev glossary

Answers & ConfidenceAbstention

Uncertainty

The honest "I'm not sure" signal that lets a system defer instead of guessing, exposed through probabilities and confidence.
Category
Answers & Confidence
Also known as
Abstention
Reading time
1 min
Related terms
5
Directory entries
2
Docs
docs.typesafe.ai
Added
2026-09-24

Definition

The confidence page opens with the principle: if an intelligent system cannot express honest uncertainty, the system cannot be trusted. Calibrated probabilities are how Jev expresses it, and the middle of a Noul's range is a natural hand-off to a person.

Uncertainty is meant to be designed with, not optimized away: low confidence often means the options are ambiguous or the state lacks information, and both are signals your code can route on.

Tagsconfidencedesign

From the Jev directory

Directory entries that go deeper on this term.

How TypeSafe reports certainty, how it differs from probability, and how to use it architecturally to gate and route decisions.
Jev Practices & PatternsDocs#official#confidence#routing
Official
Why TypeSafe trains decision models with RLCD instead of RLHF: calibrated probabilities where 0.2 outcomes happen about 20% of the time, and the case for machine-to-machine automation.
Jev Guides & ArticlesDocs#official#docs#evaluation
Official

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai