Kuaishou has open-sourced Keye-VL 1.5, a large model capable of understanding videos and performing cross-modal reasoning. Compared to the previous preview version, Keye-VL 1.5 features enhanced ...