# CG-MLLM

CG-MLLM is a 3D multimodal large language model (3D MLLM) built upon Qwen3-VL and Hunyuan3D-2.1 VAE for unified 3D understanding, 3D captioning, and high-resolution 3D content generation.

- Authors: Junming Huang, Chi Wang, Letian Li, Guangkai Xu, Donglin Huang, Hao Chen, Qiang Dai, Weiwei Xu
- Venue: ICML 2026
- arXiv: 2601.21798 — https://arxiv.org/abs/2601.21798
- Project: https://dreaming-huang.github.io/CG-MLLM-page/ (https://cv.jream.top/CG-MLLM-page/)
- Poster/ICML: https://icml.cc/virtual/2026/poster/63909
- Code: https://github.com/dreaming-huang/CG-MLLM
- Checkpoint: https://huggingface.co/JreamH/CGMLLM