A Study on a License Compatibility Determination Model Using LLM-Based Clause Understanding and Attribute Extraction

As OSS adoption grows, license conflicts emerge as critical legal and technical challenges.Existing compliance tools mainly focus on license identification and still rely on manual expert review for compatibility decisions.In practice, compatibility depends not only on license clauses but also on operational context, which makes automation difficult.We propose a structured approach in which an LLM internalizes licensing rules without an external rule engine, predicts legal/technical/contextual attributes as ternary vectors (+1, 0, -1), and directly decides compatibility given the context.The input includes full license texts and integration/distribution details, and the model outputs sentence-level rationales to ensure explainability.These results indicate that combining legal reasoning and contextual awareness in LLMs can enable compatibility judgments without explicit rule engines, providing a foundation for automated OSS license compliance.