A central challenge in multimodal in-context learning (MICL) arises from tasks that are ill-defined without context and require uncovering the underlying rul…
机构:Microsoft
来源:arXiv 2610.41754 | AI4Papers 论文推荐平台