User repository on GitHub
- Forks
- 8
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Models & Reimplementations
OmniJev (tinnel123666888)
An omni-modal Jev: a System One decision model that fine-tunes Qwen3.5 vision-language backbones with LoRA and decision heads, returning calibrated probabilities for typed questions about images, video, screens, audio, and robot scenes in one forward pass.
- Stars
- 164
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- tinnel123666888
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open source
Highlights
- Fine-tunes Qwen3.5 vision-language backbones with LoRA, decision heads, and an ordinal head.
- Returns a probability distribution per typed question in one forward pass with zero generated tokens.
- v1.1 adds multi-image training and inference, three sizes, an SFT baseline, and question-type scores.
- Can answer none of the above; the earlier v0.8 release was a 4B Qwen3.5-4B model.
- From Beijing Zhongguancun Academy, the CAS Institute of Automation, and Zevo; weights on Hugging Face.
Watch out
README reviewed 2026-10-17; the training manifest is still being audited, the earlier 270,000-record figure is not an exact count for v1.1, released demonstrations are offline replays, and closed-loop gameplay and robot control are not established.
Questions about this entry
Short answers drawn from the linked project or page.
An omni-modal Jev: a System One decision model covering images, video, phone and computer screens, games, board games, audio spectrograms, text, and robot scenes, answering typed questions with calibrated probabilities instead of generated text.
The README says you give it a state (images, video frames, screenshots, a camera feed, optional context) and a set of typed questions, and it returns a probability distribution for every question in one forward pass with zero generated tokens. It can also answer none of the above.
It fine-tunes Qwen3.5 vision-language backbones with LoRA, decision heads, and an ordinal head. The v1.1 release adds multi-image training and inference, three model sizes, an SFT baseline, and dataset and question-type scores.
Not reliably yet. The README says the released demonstrations are offline replays and that reliable closed-loop gameplay and robot control have not been established, so treat the robot and game coverage as answered questions rather than proven autonomous control.
Weights are on Hugging Face under tinnel123/OmniJev, and the README badge says Apache-2.0. GitHub metadata lists the licence as NOASSERTION, so check the repository and model card before reuse.
The README credits Beijing Zhongguancun Academy, the Institute of Automation at the Chinese Academy of Sciences, and Zevo, with a project website at omnijev.net that includes an online try-it page.
The current training manifest is still being audited, and the earlier 270,000-record figure is not an exact count for this release. The digest excerpt shows no independent benchmark results, and the demonstrations are offline replays rather than live closed-loop runs.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Happy JevDay: images, box, longer context, and an opt-in to find jaggedness
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
X: First among models under 10B on the Decision Index v0.2.1. Liquid AI's d1-3B is built from LFM2.5-VL-3B and…
X: The classic hotdog classifier as a real-time vision decision. Fayaz Ahmed rebuilt the not hotdog app on…
Other listings of this project
Forks and copies of this project in the directory. This entry is the canonical one.
Multimodal finite-choice decision interface and MuJoCo robotic-arm workbench: dual-camera images and state go to a self-hosted Qwen3.5-9B with a corrective head, which returns one typed next-skill choice without generating text.
Jev Models & Reimplementations#community#robotics#multimodal
Community
shapsider/OmniJevPython
shapsider/OmniJevPythonMore 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-visionPythonMapika's vision decider: decider-2b v5 text weights transplanted into the Qwen3.5-2B vision-language model, tuned so an image plus a lettered question returns a calibrated probability at one answer slot with no generation. Breakout from pixels scores 41, above its teacher's 22.
Jev Models & Reimplementations#community#open-models#local
Community
The vision sibling: the same text-trained System 1 head on Qwen3.8-27B's image encoder, taking images zero-shot, ranked first of 20 vision decision models on the community Jev Decision Index 0.3.
Jev Models & Reimplementations#community#open-models#local
Community
A small, fast decision model — a system one model — for agentic workflows from Strands Labs: it picks between options or rates on a scale faster than an LLM and returns a calibrated confidence with every decision.
Jev Models & Reimplementations#community#github#python
Community
strands-labs/strands-deciderPython
strands-labs/strands-deciderPythonA multimodal decision model that scores text, image, and video candidates with a shared head on Qwen3.5, plus SFT and experimental RLCD training code.
Jev Models & Reimplementations#community#python#open-models
Community
Liuziyu77/ValenJupyter Notebook
Liuziyu77/ValenJupyter NotebookAn independent open-source pipeline that gives Jev vision: it detects and tracks objects in images, video, and RGB-D camera feeds, then asks Jev a typed question about each relevant object and prints the chosen answer with its confidence.
Jev Tools & CLIs#community#github#python
Community
CharlesFeng0314/JEV_seesPython
CharlesFeng0314/JEV_seesPythonFrom 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.