Research on Metadata Standards for AI Models

Abstract

With the rapid advancement of artificial intelligence technology, the standardization and structured management of AI models have become increasingly important. However, the fragmentation of metadata standards severely compromises the interpretability, interoperability, and reusability of AI models. This study begins with a comparative analysis of existing metadata standards and examines the current application of model metadata across major AI model repositories. The analysis reveals several critical issues in current practices, including inconsistencies in metadata structures and a lack of semantic alignment. In response, this paper proposes an upper-level metadata ontology framework to support the structured and semantic description of AI models, providing a theoretical foundation for the future design of metadata interoperability mechanisms. Although the case study is limited in sample size, it offers an empirical basis for subsequent refinement and extension. Future work will focus on expanding the sample size and validating the framework in more diverse application scenarios.