JevMade hello@JevMade.com
← Back to guides

JevMade field notes / Model benchmark report

openJev-verdict-2.0

A local-model benchmark report that separates two generations, records inference fixes, and compares accuracy, Brier score, two calibration channels, latency, and perturbation flips.

Original by Heman10x-NGUEvaluationGitHub repositorySource reviewed

Before you dive in

What you’ll find in the original

  1. Keep model generations and evaluation datasets separate; similarly named checkpoints need not support the same claims.
  2. Publish test receipts and explain inference fixes such as missing calibrator loading before comparing versions.
  3. Include simple baselines: the repository's TF-IDF logistic regression was faster and had zero option flips despite lower reported accuracy.
Worth knowing

The 2,000-decision enterprise set and 77.10% accuracy are author-reported. The 1.44% ECE belongs to a separate correctness head, not the choice distribution, whose reported ECE is 15.13%.