What it is
A training library for building a small model that takes a piece of text and a list of N text options, returning one probability per option in a single pass. For developers exploring Jev-like scoring.
How it uses Jev
Jevlike is an independent starter model with the same input/output shape as Jev. It is not Jev itself. The README does not describe how Jev is used; it only notes that Jev is TypeSafe's commercial model for this task.
Primitives:choice
Technique worth stealing
Option-attention: each option becomes a query that attends to context tokens, then a shared dot product scores each option.
Try it
Run the Quickstart commands: uv venv, install, generate synthetic data, train, evaluate, predict.
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-centricyes0.78
- Shows a System One patternno0.18
- Handles uncertaintyno0.03
- Measuredno0.07
- Runnableno0.19
- Worth recommendingno0.26
- Model replicayes0.97
- Problem scopescore on a 0–2 scale0.81
- About Jevyes0.92
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.