"AgenticEvolution-22k: A Scientometric Dataset of the Shift from Chatbots to Autonomous Security (2018\u20132025)"

"This dataset contains metadata and abstracts for 22,416 research papers (2018\u20132025) documenting the evolution of artificial intelligence security. The corpus was curated via the Semantic Scholar API to capture the structural transition from perturbation-based adversarial machine learning to Large Language Model (LLM) alignment and autonomous agentic security.The primary contribution of this repository is the providing of a \"frozen\" snapshot of the field's publication trajectory, enabling reproducible trend analysis. Accompanying the dataset is a Python-based analytical pipeline that utilizes Non-negative Matrix Factorization (NMF) for unsupervised topic discovery.The provided code includes a multi-seed robustness check to validate topic stability and a specialized hierarchical \"drill-down\" modeling function. This allows researchers to quantify the emerging dominance of Autonomous Vulnerability Discovery within the broader AI Agent category, a finding central to the \"Era of Agency\" paradigm. This dataset serves as a foundational resource for scientometric research in cybersecurity and the study of autonomous systems defense."