LLMs have shown strong in-context learning (ICL) capability, but extending it to Multimodal LLMs (MLLMs) remains challenging. Prior multimodal ICL methods of…
机构:伯克利
来源:arXiv 2610.30592 | AI4Papers 论文推荐平台