International Journal of Science, Engineering and TechnologyAuthors: Chintu Kodanda Ramu, Dr.Pankaj Khairnar
Abstract: Artificial Intelligence (AI) has advanced rapidly with the development of transformer-based large language models capable of understanding and generating human language. However, traditional language models mainly process textual information and fail to integrate other forms of data such as images and speech. Human communication naturally combines multiple modalities including text, visual perception, and sound. This limitation has encouraged the development of Multimodal Large Language Models (MLLMs), which integrate text, image, and speech understanding within a unified framework. This paper examines multimodal learning approaches, transformer architectures, and multimodal fusion strategies used in modern AI systems. The study highlights how multimodal systems improve contextual understanding, emotion recognition, and human-computer interaction compared to unimodal systems. Experimental observations show that transformer-based multimodal architectures provide improved accuracy and adaptability. The paper also discusses key challenges including computational complexity, data alignment, and scalability. The findings indicate that multimodal large language models represent a major step toward building intelligent systems capable of human-like understanding.
DOI: https://doi.org/10.5281/zenodo.20095914
