Incorporating Large Language Models enhances multispectral pedestrian detection by understanding complementary information and intervening in modality bias.
