What you can build with it

What it does

The repository offers open decision models for Jev-style Choice, Score and Noul workflows, including separate Verdict and Verdict 2.0 models.

Evidence 1 ↗ · Evidence 2 ↗

How Jev fits

Verdict uses ModernBERT and GLiClass; Verdict 2.0 targets typed software workflows. These are the project’s models, with Jev used as a comparison baseline.

Evidence 1 ↗ · Evidence 2 ↗ · Evidence 3 ↗

Getting started

Serve webgpu-demo with Python’s HTTP server and open the documented browser playground; model artifacts and evaluation receipts are linked separately.

Evidence 1 ↗ · Evidence 2 ↗ · Evidence 3 ↗

Scope and limits

The two models have different evaluations. The Verdict 2.0 table cites Jev’s baseline from Laya’s evaluation suite, so its headline is not a fresh universal comparison.

Evidence 1 ↗ · Evidence 2 ↗

Repository description

Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)

Project facts

Relationship to Jev
Local alternative
Evidence
Documented
Language
Python
License
NOASSERTION
Origin
Original repository
Repository status
Not archived
Created
2026-09-19
Last code update
2026-09-20
GitHub stars
292
Evidence checked
Sep 30, 2026
Metadata checked
Sep 29, 2026
Check status
Up to date

Stars measure the whole repository, including work unrelated to Jev.

Evidence and scope

Documented records the linked documentation or source. JevHunt has not independently run or benchmarked this project.

Evidence commit: bff28567cff463b833bf044f351a8b7945d53e07

Discovered through: awesome-jev.

How we review →

Suggest a correction · Claim this listing · Submit a project