Edge of arXiv 2025: bibliometrics, themes, time trends, and networks

This study conducts a comprehensive bibliometric, thematic, temporal, and network analysis of 2000 edge computing preprints published on arXiv during 2025. Drawing from a corpus authored by 8683 researchers across 124 categories, the analysis reveals a highly collaborative field with an average of 4.86 authors per paper. Thematic modelling identifies 10 core topics, led by energy-efficient computing (659 weighted occurrences), data management frameworks (608), and AI model deployment (509), while research types emphasise systems (28.2%), machine learning (20.0%), and theory (19.2%). Temporal patterns reveal consistent growth, averaging 333 papers per month, with peaks in July (434) and October (415), likely influenced by conference cycles. Network analysis reveals 1074 communities with a modularity of 0.847, highlighting specialised clusters in federated learning and AI at the edge, although security remains underrepresented (3.4%). The field demonstrates strong AI integration (91.25%) and identifies 292 emerging topics, signalling a rapid evolution toward sustainable, quantum-enhanced, and neuromorphic paradigms. Findings underscore gaps in security, real-world evaluation, and sustainability, while proposing directions for interdisciplinary advancement.