What it is
A Python toolkit and CLINC150 study that turns Jev's calibrated probabilities into certified routing thresholds with finite-sample guarantees, and audits them on live traffic using prediction-powered inference. For teams deploying Jev-based routers who need provable risk bounds.
How it uses Jev
Jev answers two typed questions per query: a `choice` for the intent and a `noul` for scope. Conformal risk control uses these to set thresholds with a bound on silently misrouted queries. The certificate is audited with PPI++ using few hand labels.
Primitives:choicenoul
Technique worth stealing
Conformal risk control turns calibrated probabilities into certified routing thresholds; PPI++ audits them with few labels.
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.93
- Shows a System One patternunclear0.41
- Handles uncertaintyunclear0.47
- Measuredyes0.93
- Runnableno0.21
- Worth recommendingunclear0.45
- Model replicano0.03
- Problem scopescore on a 0–2 scale1.17
- About Jevyes0.99
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.