3D-MoE: A Mixture-of-Experts Multi-modal LLM for 3D Vision and Pose Diffusion via Rectified Flow
This work addresses 3D vision and embodied task planning for AI systems, representing an incremental advancement by adapting existing methods to this domain.
The paper tackles the challenge of 3D vision and spatial reasoning by proposing 3D-MoE, a multi-modal LLM that converts existing models into mixture-of-experts and adds a diffusion head for task planning, achieving improved performance with fewer activated parameters.
3D vision and spatial reasoning have long been recognized as preferable for accurately perceiving our three-dimensional world, especially when compared with traditional visual reasoning based on 2D images. Due to the difficulties in collecting high-quality 3D data, research in this area has only recently gained momentum. With the advent of powerful large language models (LLMs), multi-modal LLMs for 3D vision have been developed over the past few years. However, most of these models focus primarily on the vision encoder for 3D data. In this paper, we propose converting existing densely activated LLMs into mixture-of-experts (MoE) models, which have proven effective for multi-modal data processing. In addition to leveraging these models' instruction-following capabilities, we further enable embodied task planning by attaching a diffusion head, Pose-DiT, that employs a novel rectified flow diffusion scheduler. Experimental results on 3D question answering and task-planning tasks demonstrate that our 3D-MoE framework achieves improved performance with fewer activated parameters.