jevbooks

← All projects

Visual-Jev

jiangxiluning/Visual-Jev

Open visual semantic decision model with typed option probabilities.

Agent Decisions93%Runner-up: Infra / SDKs / Integrations

What it is

SemIf (formerly OpenJev) is an independent open-source project that reproduces Jev's interface pattern using open models. It reads typed option probabilities directly from a 4B vision-language model, enabling runtime-defined semantic decisions without text generation.

How it uses Jev

Jev is not used; SemIf is an independent reproduction of Jev's interface pattern. It takes a state, question, and options, and returns option logits as probabilities via one forward pass. Optional image input is fused through the native processor.

Primitives:choice

Technique worth stealing

Direct readout of option logits from a frozen vision-language model, avoiding autoregressive generation.

Try it

pip install -e '.[test]' then run semif-score with --mode direct.

View on GitHub

judged by Jevjev-1.13.0

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-centricyes0.62
  • Shows a System One patternno0.34
  • Handles uncertaintyno0.04
  • Measuredno0.11
  • Runnableyes0.76
  • Worth recommendingno0.21
  • Model replicayes0.91
  • Problem scopescore on a 0–2 scale0.64
  • About Jevyes0.93

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 22 Sept 2026.