What it is
SemIf is an independent open-model research baseline for direct typed option scoring. It reproduces the interface pattern of runtime-defined semantic decisions, not Jev's undisclosed model or training. For developers running semantic decisions on a local GPU.
How it uses Jev
The README does not describe using Jev. SemIf is independent and not affiliated with Jev or TypeSafe. It reads typed option probabilities directly from open models via native option logits, with no answer sentence, JSON repair, or decoding loop.
Primitives:choice
Technique worth stealing
Read typed option probabilities directly from native option logits in one forward pass, avoiding generated text.
Try it
pip install -e '.[test]' then run semif-score with --mode direct on examples/decisions.jsonl.
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricno0.37
- Shows a System One patternunclear0.43
- Handles uncertaintyno0.04
- Measuredno0.11
- Runnableunclear0.55
- Worth recommendingno0.12
- Model replicayes0.88
- Problem scopescore on a 0–2 scale1.09
- About Jevyes0.86
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.