Multimodal LLMs with well-designed prompts can partially match human judgments of driving-scene risk factors, but their performance depends heavily on prompt design and remains unreliable for safety-critical detection.
How ai from automated driving systems can contribute to the assessment of human driving behavior,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Understanding Driving Risks using Large Language Models: Toward Elderly Driver Assessment
Multimodal LLMs with well-designed prompts can partially match human judgments of driving-scene risk factors, but their performance depends heavily on prompt design and remains unreliable for safety-critical detection.