open-jev (JoshuaSP)
Typed JSON inference with DiffusionGemma, with Every and Jev benchmark results
How it uses Jev
The goal is to explore how close an open diffusion model can get to TypeSafe AI's Jev on structured judgments. This is an independent experiment, not a reproduction of Jev's architecture, training, calibrated probabilities, or published workflow scores.
Getting started
Requires Python 3.11+, a Modal account, and a Hugging Face token with access to the model. Accept the model's access terms on Hugging Face first.
Source: community catalog. Details and metrics reflect the published community snapshot and are not a performance endorsement. Documentation details extracted from GitHub · the README.