velrim-eval
v0.0.2
Published
Thin CLI to score document-extraction adapters on a 3-state golden set. Golden in -> per-field F1 + reliability curve out -> CI exit code. Publishes in full with the benchmark results.
Readme
velrim-eval
A thin CLI to score document-extraction adapters on a 3-state golden set: golden in, per-field F1 and a reliability curve out, a CI exit code at the end. Not a platform.
The full source, the benchmark pre-registration, and the frozen evaluation materials live at github.com/velrimhq/velrim-eval. The CLI publishes here in full together with the benchmark results.
Apache-2.0.
