TheoLeeCJ/openjev
- Source
- GitHub
- First trending
- Category
- Model inference
- GitHub stars
- 824
- Main language
- Python
This page introduces an external open-source repository. It is not an HDATF product.
What it does
An open-model experiment that reads typed option probabilities directly instead of generating answer text. It explores Jev's interface pattern rather than reproducing its undisclosed model or training.
How it helps ATF
Harness can compare this approach for small routing and retry decisions that do not require generated prose.
License
MIT Permissive. Commercial use and changes are allowed if the copyright notice is kept.
More in this category
- anthropics/uplifting-biomolecular-modeling
A reference collection of inference optimization kits for protein and genomics machine-learning tools at pinned upstream versions. The release is not maintained and does not accept contributions. - PrismML-Eng/Bonsai-demo
A demo for running Bonsai language models locally on GPUs or CPUs. Its Bonsai 2 demo includes vision inputs, tool calls and MCP server support. - hr98w/jev-visual
An educational Apple Silicon experiment using Qwen3.5 and MLX to answer multiple questions about one image by scoring candidate outputs. It includes a local browser interface, CLI and HTTP API. - vinnylarouge/jevlike
Jevlike is an independent starter model that reads a text and a changing list of text options and returns one probability per option in a single pass. Demos apply the same approach to Doom controller buttons and chess moves. - featherless-ai/simple-jev
Uses compatible open models for structured classification and scoring without training a separate classifier head. The server constructs JSON choices, rubric scores, or truth and support judgments from next-token logits rather than generated JSON text.
Only repositories in the ranked Trendshift lists are included, and the lists are used only to find candidates. We do not copy their ranks. Descriptions, licenses and star counts come from each GitHub repository. The notes are our own reading. We have not tested these projects, and a place on a trending list does not prove quality.