Using Large Multimodal Models to Extract Knowledge Components for Knowledge Tracing from Multimedia Question Information