Announcement_15

We propose MMSearch-Plus — a challenging MLLM agent benchmark that requires high-level multimodal understanding and tool-use capability through iterative image-text retrieval and cross-validation under retrieval noise. Released benchmark dataset is available here.