MoIIE: Mixture of Intra- and Inter-Modality Experts for Large Vision Language Models
This work addresses computational efficiency for LVLMs, offering a novel MoE architecture that improves parameter efficiency while maintaining performance, which is incremental but impactful for multi-modal AI applications.
The paper tackles the challenge of efficiently applying Mixture of Experts (MoE) to Large Vision-Language Models (LVLMs) by proposing MoIIE, which uses modality-guided routing to learn intra- and inter-modality features, resulting in models with 5.5B and 11.3B activated parameters matching or surpassing existing open-source MoE-based multi-modal models with more parameters.
Large Vision-Language Models (LVLMs) have demonstrated remarkable performance across multi-modal tasks by scaling model size and training data. However, these dense LVLMs incur significant computational costs and motivate the exploration of sparse Mixture of Experts (MoE) architectures. While MoE improve parameter efficiency, effectively applying MoE to simultaneously model modality-specific features and cross-modal associations in LVLMs remains challenging. In this work, we propose to incorporate Mixture of Intra- and Inter-Modality Experts (MoIIE) to LVLMs. For each token, expert routing is guided by its modality, directing tokens to their respective intra-modality experts as well as a shared pool of inter-modality experts, enabling the model to jointly learn rich intra-modal features and cross-modal interactions. We further introduce an effective and straightforward two-stage training strategy, which facilitates the direct activation of both MoE and multi-modal capabilities. Extensive experiments across different data scales and LLM backbone demonstrate the effectiveness, efficiency and generality of our approach. Notably, our MoIIE models with 5.5B and 11.3B activated parameters match or even surpass the performance of existing advanced open-source MoE-LLMs based multi-modal models that involve more activated parameters. The code is available at https://github.com/AlenjandroWang/MoIIE.