Incongruence-Aware Evaluation of Multimodal Speech Emotion Recognition in Audio Large Language Models | ArxivCSExplorer