Benchmark RadarObsDriveBench
ObsDriveBench evaluates multimodal understanding in autonomous driving under adverse weather, covering observability awareness, spatial reliability, and risk-aware decision-making with multiple-choice tasks over camera, LiDAR, and radar inputs.
- Released
- 2026-07-26
- Readiness
- Runnable
- Primary field
- Transport & Logistics
Why it matters
Existing benchmarks rarely assess vision-language models under real-world adverse conditions with multimodal inputs. ObsDriveBench targets this gap by providing a fine-grained diagnosis of model behavior when observations are unreliable, supporting safer autonomous driving systems.
Motivation
Autonomous driving under adverse weather remains a critical challenge, yet existing vision-language benchmarks mainly evaluate under standard conditions, synthetic corruptions, or single modality.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.