worldbench/DriveBench

[ICCV 2025] Are VLMs Ready for Autonomous Driving? An Empirical Study from the Reliability, Data, and Metric Perspectives

43
/ 100
Emerging

DriveBench is a dataset for evaluating how well Vision-Language Models (VLMs) understand complex driving scenarios. It takes in images and text-based questions about driving situations and outputs answers that reveal if the VLM truly understands the visual context, especially under challenging conditions. Autonomous driving researchers and engineers can use this to rigorously test and improve the reliability of their VLM-powered systems.

232 stars.

Use this if you are developing or evaluating AI systems for autonomous vehicles and need to rigorously test how reliably Vision-Language Models interpret driving scenes under various conditions, including degraded visual input.

Not ideal if you are looking for a dataset to train general-purpose Vision-Language Models outside of the autonomous driving domain.

autonomous-driving vehicle-AI VLM-evaluation perception-systems robotics-safety
No Package No Dependents
Maintenance 6 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 11 / 25

How are scores calculated?

Stars

232

Forks

15

Language

Python

License

Apache-2.0

Last pushed

Dec 12, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/worldbench/DriveBench"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.