[ICLR 2025] VL-ICL Bench: The Devil in the Details of Multimodal In-Context Learning - View it on GitHub