Explore / Library / SDK / robo-jev — Physical Judgment Model

robo-jev — Physical Judgment Model

Self-hostTerminal

An independent 2B model for text-based physical judgments between robot planning and control.

by admin · listed 5 days ago

⚡ What System1 does here

It sits between task planning and low-level motion control, choosing actions such as continuing, slowing or stopping and judging collision, tipping, spilling or part-seating conditions. Each question uses a fixed prompt and reads answer-position probabilities into choice, noul or score outputs. It does not provide visual perception or replace motor controllers.

About

STEERIX's open-source robot decision model and simulation-based training and evaluation toolkit. The current README identifies v0.12, fine-tuned from Qwen3.5-2B. It reads textual sensor-state descriptions and returns probabilities for physical judgments and action choices. It is an independent model; TypeSafe Jev is one benchmark comparator. Evidence and limits: An independent research model fine-tuned from Qwen3.5-2B, not official TypeSafe Jev. The author reports weak reliability outside simulation and susceptibility to incorrect planner hints, excluding safety-system use. The chart is a project-run simulation benchmark, not real-robot reliability evidence. No public video was found. This review checked public documentation, code and media; it did not rerun simulations or test physical hardware. Original project images and demonstrations (with attribution):

Project-run v0.12 simulation benchmark ↗

Sources: https://github.com/STEERIX-home/robo-jev/blob/main/README.md https://github.com/STEERIX-home/robo-jev/blob/main/src/robo_jev/hf/robo_jev_decide.py https://github.com/STEERIX-home/robo-jev/blob/main/docs/method.md https://huggingface.co/Steerix-org/robo-jev-2b https://robo-jev.dev/

🧪 Community test reports

0 reports · 0 passed on the real model

No one has tested this yet. Be the first.

💬 Discussions

New post →

No discussions yet.