Classification and Generation of Images Containing Armwrestling Techniques

Large language models (LLMs) are often incapable of performing highly specialized image classification and generation tasks (for example, in the domain of armwrestling techniques). This paper investigates the training of artificial intelligence models for the tasks of classification and generation of images containing armwrestling techniques. The paper also includes a comparison of the obtained results with existing multimodal LLMs.