AI benchmark for self-driving cars now measures what different viewpoints can see

Researchers built a test for AI systems used in autonomous vehicles that measures performance across three different perspectives: what the car sees, what roadside sensors see, and what both see together. Until now, most tests only measured what the car itself could perceive, missing large gaps in how well AI reasons when it gets information from multiple sources.

Autonomous vehicles that talk to infrastructure and other vehicles are coming. The gap between single-car AI and multi-car AI is invisible until you measure it. This benchmark reveals that AI systems actually struggle with data from multiple viewpoints, not because they lack information, but because they can't align what different cameras and sensors are seeing into a coherent picture. That's a structural problem, not a data problem.