User repository on GitHub
Image-native typed decisions with shared visual encoding and Qwen3-VL
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Models & Reimplementations
Visual Jev
An open, local attempt to bring Jev-style typed decisions to images: a Qwen3-VL backbone returns numerical distributions for independent choice, score and noul questions without generating answer text.
- Stars
- 3
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- andrueandersoncs
- Last verified
- Oct 2026
Used inJev for Software & DevTools
Highlights
- Image-native typed decisions: images plus optional context in, numerical distributions for choice, score and noul questions out — no answer text.
- Built on a Qwen3-VL backbone with isolated packed question branches, a learned pointer head, language-backbone LoRA and held-out temperature calibration.
- Missing, experimental, incompatible or checksum-mismatched weights fail closed: the server never substitutes baseline or random weights.
- The candidate-token Qwen path is kept only as an explicit uncalibrated baseline, separate from the calibrated production path.
- The promoted checkpoint is served with uv run visual-jev --registry models --model visual-jev-public-mps-bf16-2026-03-v3.
Quickstart
bash
uv sync --extra dev
npm --prefix web ci
npm --prefix web run build
uv run visual-jev --registry models --model visual-jev-public-mps-bf16-2026-03-v3Watch out
Copy reviewed against the repo README on 2026-10-15; the README is short and the operational detail lives in docs/OPERATIONS.md, which was not part of the capture. GitHub metadata declares no license and the README reports no accuracy numbers.
Questions about this entry
Short answers drawn from the linked project or page.
An open, local project applying the Jev pattern to images: it accepts one or more images with optional structured context and returns numerical distributions for independent choice, score and noul questions — without generating answer text.
It is a community reimplementation of the typed-decision idea over a different modality: decisions over images directly, rather than over text descriptions of images. The README describes it as open and local, using a Qwen3-VL backbone.
Sync dependencies with uv sync --extra dev, build the web package with npm --prefix web ci and npm --prefix web run build, then start it with uv run visual-jev --registry models --model visual-jev-public-mps-bf16-2026-03-v3.
A promoted checkpoint must exist at models/visual-jev-public-mps-bf16-2026-03-v3. Missing, experimental, incompatible or checksum-mismatched weights fail closed — the server never falls back to baseline or random weights.
The production path returns calibrated distributions from the typed question branches. A candidate-token Qwen path remains available only as an explicit uncalibrated baseline, the README says.
The architecture uses held-out temperature calibration, and its operations guide covers untouched calibration and test evaluation, controlled head-only and candidate baselines, and an exact-input comparison against pinned OpenJev.
It is credited to Andrue Anderson in the madewithjev listing; the repo is Python, created 2026-09-19 and last pushed 2026-09-20. No license is declared in the GitHub metadata, and no benchmark results appear in the README.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
More Jev resources like this
An independent fork of Laya that replaces its text encoder with a small vision-language model, giving calibrated typed decisions over an image plus text in one forward pass from a 201M-parameter SmolVLM-256M backbone.
Jev Models & Reimplementations#community#catalog-import#cobanov
Community
r33drichards/laya-visionPython
r33drichards/laya-visionPythonA Japanese System One decision model built on modernbert-ja-310m: send text and typed questions and get typed answers with probabilities from one forward pass, no text generation, at 314.6M parameters that also run on a CPU.
Jev Models & Reimplementations#community#catalog-import#cobanov
Community
hiroki-abe-58/sokudanPython
hiroki-abe-58/sokudanPythonAn open-source, Jev-compatible System One model: Dex answers choice, score and yes/no questions over text in a single forward pass from an Ettin-1B encoder, and speaks the same /v1/systemone shape as Jev.
Jev Models & Reimplementations#community#catalog-import#madewithjev
Community
canadaornaw/dexPython
canadaornaw/dexPythonGestalt Labs' Jeff is a LoRA adapter on Qwen3-4B-Instruct-2507 that answers Jev-style choice, noul and score questions with labels and probabilities instead of prose, runs locally, and ships a train-your-own guide; on 9,730 fact-checking examples it trails Jev 1.13.0 on accuracy (0.8183 vs 0.8283).
Jev Models & Reimplementations#community#catalog-import#madewithjev
Community
A CPU port of SemIf with a web UI: JEV-CPU loads Qwen3-0.6B in float32 and reads a decision from its option logits in about a second, across eight domains from support to incident severity, with no GPU.
Jev Models & Reimplementations#community#catalog-import#madewithjev
Community
Zefan Cai's LoRA adapter and scalar decision head for Qwen3.5-9B score caller-supplied candidates for choice, yes/no and score answers without generating text, served over a local /v1/systemone endpoint.
Jev Models & Reimplementations#community#catalog-import#madewithjev
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.