CITATION.cff

cff-version: 1.2.0
message: "If you use Journeyman in your work, please cite it as below."
title: "Journeyman: a process-quality benchmark for LLM agents"
abstract: >-
  Journeyman grades how an agent worked rather than whether it passed. It
  drops an agent into small simulated jobs and scores the walk: whether it
  kept hitting the same wall, whether it stopped when the job was done,
  whether it could price what it could not reach, whether it bought a
  planted false story. Judged axes are scored by a model judge, and no
  judge is trusted on reputation — every judge sits a published,
  calibrated exam against labelled real records before it may score.
type: software
authors:
  - family-names: "Barlık"
    given-names: "Onur"
    affiliation: "Codechu"
doi: 10.5281/zenodo.22085820
repository-code: "https://github.com/codechu/journeyman"
url: "https://pypi.org/project/journeyman-bench/"
license: Apache-2.0
version: 0.4.0
date-released: "2026-08-24"
keywords:
  - benchmark
  - agent evaluation
  - process quality
  - LLM agents
  - LLM-as-a-judge