Disentangling Pictorial Cue Understanding from Language Bias in VLMs via Depth Ordering Task figure
AlphaXiv 中文概览(可滚动查看)