NaViL: Rethinking Scaling Properties of Native Multimodal Large Language Models under Data Constraints
The Chinese University of Hong Kong · Shanghai Artificial Intelligence Laboratory · Xiamen University · Shanghai AI Laboratory · Sensetime · Xi'an Jiaotong University · Nanjing University · University of Chinese Academy of Sciences · Chinese Academy of Sciences · SenseTime · Tsinghua University
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
Compositional training has been the de-facto paradigm in existing Multimodal Large Language Models (MLLMs), where pre-trained vision encoders are connected with pre-trained LLMs through continuous multimodal pre-training. However, the multimodal scaling property of this paradigm remains difficult to explore due to the separated training. In this paper, we focus on the native training of MLLMs in an end-to-end manner and systematically study its design space and scaling property under a practical setting, i.e., data constraint. Through careful study of various choices in MLLM, we obtain the optimal meta-architecture that best balances performance and training cost. After that, we further explore the scaling properties of the native MLLM and indicate the positively correlated scaling relationship between visual encoders and LLMs. Based on these findings, we propose a native MLLM called NaViL, combined with a simple and cost-effective recipe. Experimental results on 14 multimodal benchmarks confirm the competitive performance of NaViL against existing MLLMs. Besides that, our findings and results provide in-depth insights for the future study of native MLLMs.