Translated data: Tencent AI Lab collaborates with the team from the Chinese University of Hong Kong to introduce UniRepLKNet, challenging the dominance of Transformers in the multi-modal field. This large-kernel CNN architecture excels in tasks such as point clouds, audio, and video without altering the model structure. UniRepLKNet surpasses Transformers in tasks like ImageNet, COCO, and ADE20K, demonstrating the potential of large-kernel CNNs in multi-modal applications.