---
title: RoboJev · Did it work?
colorFrom: green
colorTo: gray
sdk: static
app_file: index.html
license: apache-2.0
short_description: Robot hours, the model, and the results behind the launch
models:
  - josefchen/RoboJev-27B
---

# RoboJev launch writeup

One-page summary with the released model, launch charts and input diagram, full recorded benchmark comparison and the assumptions that went into the robot-hour scenario.

[Model](https://huggingface.co/josefchen/RoboJev-27B) · [Interactive demo](https://robo-jev.com/demo) · [Data explorer](https://robo-jev.com/data)

## Research paper

**RoboJev: An Open Vision-Language Judge for Robot Success Detection**
Josef Chen · Independent Researcher

[arXiv](https://arxiv.org/abs/2610.33718) · [PDF](https://huggingface.co/spaces/josefchen/RoboJev/resolve/main/paper/RoboJev.pdf) · [LaTeX source](https://huggingface.co/josefchen/RoboJev-27B/resolve/main/paper/RoboJev-arxiv-source.zip) · [Reproducibility bundle](https://huggingface.co/josefchen/RoboJev-27B/resolve/main/paper/RoboJev-reproducibility.zip) · [BibTeX](https://huggingface.co/josefchen/RoboJev-27B/resolve/main/paper/CITATION.bib)

arXiv preprint [2610.33718](https://arxiv.org/abs/2610.33718). The paper describes the evaluated checkpoint at `e089a9b69a365744b80765057d86583647af95a9`.

The Space code and existing launch assets retain their declared licences. Adding the manuscript does not change its separate publication rights; third-party figure inputs retain their credited licences.

The main website is [robo-jev.com](https://robo-jev.com). This Space remains a launch-page mirror; model weights, datasets and GPU scoring stay on Hugging Face.