Tag
This paper examines how the order of conflicting evidence in multimodal large language models affects judgments, revealing cross-modal evidence noncommutativity where placing perceptual evidence later increases model reliance on it.
This paper proposes Joint Speech-Text Interleaved Pretraining (JSTIP), a pretraining strategy that constructs word-level and segment-level interleaved speech-text sequences to improve ASR entity accuracy and reduce the modality gap between speech and text, showing competitive performance on domain adaptation and zero-shot speech question answering.