STRUCTURAL CLOSURE: An Execution Architecture that Resolves Recurrent Failure in LLM Agents by Sealing Blanks and Unifying the Decision Path

This project archives the research work “Structural Closure,” an execution-level architecture designed to seal decision-path contamination in LLM-based agents. The study introduces NCAF, PCM, and PVM as a unified decision-admissibility framework. The repository includes both English and Korean versions of the paper. GitHub (public release): https://github.com/NorthStarAILab/structural-closure Research type: IndepeKorean parallel-language edition is available as an OSF component (“Korean Version (Parallel-Language Edition)”). ndent, non-commercial.