MindEdit-Bench: Benchmarking Object-Level Counterfactual Spatial Reasoning in VLMs from In-the-Wild Photos figure
AlphaXiv 中文概览(可滚动查看)