jevbooks

← All projects

Verdict-open-jev

Heman10x-NGU/Verdict-open-jev

Non-autoregressive decision engine with calibrated uncertainty and WebGPU playground.

Agent Decisions35%Runner-up: Evaluation & BenchmarkingMany questions, one request

What it is

OpenJev (Verdict) is an open-source post-trained foundational decision model for structured software workflows, inspired by TypeSafe AI's Jev and RLCD. It evaluates typed questions in a single forward pass, returning choices, scores, and probabilities with calibrated confidence.

How it uses Jev

Jev is used as a benchmark audit: the README reports a TypeSafe AI Jev benchmark audit and evaluates OpenJev on 231 public JevBench tasks. The README does not specify how Jev itself is invoked or integrated in code.

Primitives:choicescorenoul

Technique worth stealing

NLI sentence templating for candidate labels and auto-calibrator with per-k scaling.

Try it

Run the in-browser WebGPU playground (see README section).

View on GitHub

judged by Jevjev-1.13.0

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-centricyes0.79
  • Shows a System One patternno0.32
  • Handles uncertaintyno0.11
  • Measuredno0.11
  • Runnableyes0.69
  • Worth recommendingno0.33
  • Model replicayes0.94
  • Problem scopescore on a 0–2 scale1.66
  • About Jevyes0.93

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.