What it does
This is an experiment about scene descriptions, not a validated driving system.
Benchmarks & research
Compare Jev's driving choices when its road description comes from perfect simulator state or a vision model.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
This is an experiment about scene descriptions, not a validated driving system.
You can test how language models make fast navigation choices without looking at camera images directly. Start by planning simulation obstacles, like sudden stalled cars or clear passing lanes, to test whether descriptive text is enough for good choices. Your developer will set up the 3D test simulator and supply an access key connecting it to the model.
During testing, the software automatically writes a short summary of the road twice a second and asks the model to pick between cruising, slowing, stopping, or overtaking. Regular computer code carries out the steering and steps in with emergency braking if a crash is imminent. Remember that this setup is a research experiment, not safe software for actual vehicles.