MLLMs Need 3D-Aware Representation Supervision for Scene Understanding
This work addresses the challenge of improving 3D scene understanding for AI systems using MLLMs, representing an incremental advancement by enhancing existing models with 3D supervision.
The paper tackles the problem of limited 3D representation capability in multimodal large language models (MLLMs) for scene understanding by proposing 3DRS, a framework that supervises MLLM visual features with knowledge from pretrained 3D foundation models, resulting in consistent performance gains across benchmarks like visual grounding, captioning, and question answering.
Recent advances in scene understanding have leveraged multimodal large language models (MLLMs) for 3D reasoning by capitalizing on their strong 2D pretraining. However, the lack of explicit 3D data during MLLM pretraining limits 3D representation capability. In this paper, we investigate the 3D-awareness of MLLMs by evaluating multi-view correspondence and reveal a strong positive correlation between the quality of 3D-aware representation and downstream task performance. Motivated by this, we propose 3DRS, a framework that enhances MLLM 3D representation learning by introducing supervision from pretrained 3D foundation models. Our approach aligns MLLM visual features with rich 3D knowledge distilled from 3D models, effectively improving scene understanding. Extensive experiments across multiple benchmarks and MLLMs -- including visual grounding, captioning, and question answering -- demonstrate consistent performance gains. Project page: https://visual-ai.github.io/3drs