User repository on GitHub
Jeva: System One Models from First Principles
- Forks
- 5
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Models & Reimplementations
Jeva — liqiiiii
System One models from first principles by Qi Li and Xinchao Wang at the National University of Singapore: give Jeva a state and typed questions and it returns candidate probabilities, a Boolean probability, or an expected rubric score, with reusable option memory.
- Stars
- 93
- Language
- Python
- License
- APACHE-2.0
- Reading time
- 2 min
- Published by
- Community
- Author
- LiQiiiii
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceRuns locally
Highlights
- Four structural properties from the README: null awareness, permutation invariance, option scalability, and reusable option memory.
- A language encoder represents the state and each option independently; a compact decision stack updates only the state against the complete option memory.
- Permutation symmetry: reordering candidates permutes their probabilities rather than changing their semantic identities; options are encoded in bounded batches.
- Install targets Linux x86_64 with an NVIDIA GPU: a CUDA 12.8.1 conda environment, PyTorch 2.9.1 requirements, and jeva-doctor --check-kernels --strict.
- GitHub: 85 stars, 5 forks, Python, Apache-2.0, created 2026-10-12; model at huggingface.co/LIQIIIII/Jeva-4B; paper arXiv 2610.18773.
Quickstart
bash
conda create -n jeva -c nvidia/label/cuda-12.8.1 python=3.12 cuda-nvcc=12.8.93 -y
conda activate jeva
git clone https://github.com/LiQiiiii/Jeva.git
cd Jeva
python -m pip install --upgrade pip
python -m pip install -r requirements.txt
jeva-doctor --check-kernels --strictWatch out
README reviewed 2026-10-17 and truncated inside the local inference example, so most of the Python API and the linked evaluation results are not in the digest. No independent benchmarks, pricing, or hardware minimums appear beyond Linux x86_64 with an NVIDIA GPU and a compatible driver.
Questions about this entry
Short answers drawn from the linked project or page.
A System One model with native design from Qi Li and Xinchao Wang at the National University of Singapore: it takes a state plus typed questions and returns candidate probabilities, a Boolean probability, or an expected rubric score.
Instead of generating text, it answers typed questions in one pass: the language encoder embeds the state and each option independently, and a compact decision stack updates only the state against the complete option memory.
Null awareness separates answer availability from competition between candidates with an explicit coverage estimate; permutation symmetry keeps probabilities tied to identities when options are reordered; option scalability avoids a fixed label vocabulary; and two-level reuse caches option tuples or vectors.
The README targets Linux x86_64 with an NVIDIA GPU: create the CUDA 12.8.1 conda environment, clone the repo, install the requirements (PyTorch 2.9.1), then run jeva-doctor --check-kernels --strict. The first request may compile kernels.
Linux x86_64, an NVIDIA GPU, and a compatible driver, per the README; the accelerated requirements pair PyTorch 2.9.1 with CUDA 12.8. A slower reference installation exists in requirements-core.txt, and runtime memory limits are documented in docs/runtime.md, which is not in the digest.
Yes: the README links a Jeva-4B model on Hugging Face, an evaluation document, a fine-tuning document, a project page, and an arXiv citation (2610.18773). The digest does not include benchmark results.
Qi Li and Xinchao Wang of the National University of Singapore; the GitHub metadata shows 85 stars, 5 forks, Python, Apache-2.0, created 2026-10-12, and a push on 2026-10-14, so it was days old at verification.
Other listings of this project
Forks and copies of this project in the directory. This entry is the canonical one.
A self-trainable 2B browser-agent decision model that speaks TypeSafe's /v1/systemone protocol, so an existing jev-ultrafast agent is repointed with a one-line base-URL change. Runs as a 1.6 GB Q4_K_M GGUF under llama.cpp at about 236 ms per decision.
Jev Models & Reimplementations#community#github#python
Community
chemany/jevaPython
chemany/jevaPythonMore Jev resources like this
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-deciderPythonAn open-source, local alternative to Jev: System-One decision models that answer the POST /v1/systemone shape — choice, yes/no, or score — for general decisions and browser-agent steps. Models ship as OpenJev-1.7B, 4B, and 8B, with a pip-installable Python package.
Jev Models & Reimplementations#community#github#python
Community
alanhuangyoo/openjevPython
alanhuangyoo/openjevPythonAn open System One decision model from Interfaze-ai: give lev a state — text, a ticket, an email, or JSON — plus yes/no, choice, or score questions and it answers all of them in one forward pass from logits, returning calibrated probabilities and zero output tokens.
Jev Models & Reimplementations#community#github#python
Community
abhinavexists/levPython
abhinavexists/levPythonA Chinese-language decision model that returns a probability for each fixed candidate option in one forward pass, with no generated text or chain of thought. It is a LoRA fine-tune of Qwen3.5-0.8B on 13,048,000 CJ-Dataset rows, benchmarked on CJ-Bench against Chinese-Jev.
Jev Models & Reimplementations#community#github#python
Community
yanqiangmiffy/chinese-decisionPython
yanqiangmiffy/chinese-decisionPythonSmall Jev-style decision models that read a state and typed questions and return a probability for every option in one forward pass without generating text; the 0.75B and 4.2B versions speak TypeSafe's Jev API format at a median of about 50 ms per request on one H100.
Jev Models & Reimplementations#community#github#decision-model
Community
jiwidi/jiwoPython
jiwidi/jiwoPythonA generation-free typed-decision engine on Sarvam's Indic models: options become lettered slots, the answer is read from those letters' logits, and one state prefill is amortised across many cheap questions. Its README says the interface pattern is TypeSafe's Jev.
Jev Models & Reimplementations#community#github#python
Community
sagar-tamang/sarvam-jevPython
sagar-tamang/sarvam-jevPythonFrom 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.