MultAttnAttrib: Training-Free Multimodal Attribution in Long Document Question Answering figure
AlphaXiv 中文概览(可滚动查看)