SPHINX-X: Scaling Data And Parameters For A Family Of Multi-modal Large Language Models

Liu Dongyang, Zhang Renrui, Qiu Longtian, Huang Siyuan, Lin Weifeng, Zhao Shitian, Geng Shijie, Lin Ziyi, Jin Peng, Zhang Kaipeng, Shao Wenqi, Xu Chao, He Conghui, He Junjun, Shao Hao, Lu Pan, Li Hongsheng, Qiao Yu, Gao Peng. Arxiv 2024

[Paper] [Code]    
Efficiency And Optimization Has Code Model Architecture Multimodal Models Tools Training Techniques

We propose SPHINX-X, an extensive Multimodality Large Language Model (MLLM) series developed upon SPHINX. To improve the architecture and training efficiency, we modify the SPHINX framework by removing redundant visual encoders, bypassing fully-padded sub-images with skip tokens, and simplifying multi-stage training into a one-stage all-in-one paradigm. To fully unleash the potential of MLLMs, we assemble a comprehensive multi-domain and multimodal dataset covering publicly available resources in language, vision, and vision-language tasks. We further enrich this collection with our curated OCR intensive and Set-of-Mark datasets, extending the diversity and generality. By training over different base LLMs including TinyLlama1.1B, InternLM2-7B, LLaMA2-13B, and Mixtral8x7B, we obtain a spectrum of MLLMs that vary in parameter size and multilingual capabilities. Comprehensive benchmarking reveals a strong correlation between the multi-modal performance with the data and parameter scales. Code and models are released at https://github.com/Alpha-VLLM/LLaMA2-Accessory

Similar Work