About the High-resolution Image Encoder (light-weight)

[ZhangWeihang99]

May I ask if the CogAgent in the paper is the same model as the open-source CogAgent9B-20241220? I haven't come across the corresponding lightweight high-resolution image encoder. It seems that CogAgent9B-20241220 is based on ChatGLM, which differs from the CogVLM described in the paper.

[sixsixcoder]

你是对的，CogAgent9B-20241220是基于GLM-4v的，论文应该是上一个版本的技术报告