Evidence in Networked Metaverse Systems: Comparability, Reproducibility, and Deployment

Abstract

Metaverse research now spans immersive networking, edge computing, digital twins, semantic communication, artificial intelligence, security, interoperability, sustainability, and deployment. The field is technically rich, yet apparently similar results are frequently produced under incompatible workloads, hardware, networks, baselines, metric semantics, and evaluation protocols. This survey therefore treats evidence rather than technology labels as the unit of synthesis. We developed a mechanism-first evidence architecture, implemented the associated validation and provenance workflow as research software, executed the workflow on a frozen 200-record corpus, and used the resulting artifacts as the empirical basis of the synthesis. The framework combines the comparability tuple C = (P, W, D, H, N, B, M, E), C0–C3 comparison classes, a multidimensional evidence vector, R0–R4 artifact-reproducibility states, denominator-aware reporting, and contrary-evidence gap falsification. The corpus contains 38 networking/edge/XR records, 21 digital-twin records, 18 semantic-communication records, 25 AI/GenAI records, 32 security/privacy records, 9 standards/interoperability records, 18 sustainability records, 12 datasets/testbeds records, and 27 foundations/applications/other records. Coder B completed and approved a second-review validation of 50 of 200 records (25%), and the approved workbook is preserved unchanged as a raw audit artifact. Because this validation used an approve/modify/reject review of pre-populated codes rather than a blind, independently generated second label vector, Cohen’s κ would not measure independent inter-rater reliability and is therefore not reported. A source-coded 21-pair validation subset yields C0 = 2 (9.5%), C1 = 6 (28.6%), C2 = 13 (61.9%), and C3 = 0; these values are explicitly bounded to that subset rather than extrapolated to the 180-pair prospective frame. All 15 closest surveys were source-verified at review level, which falsifies broad claims that resource-allocation benchmarking, sustainability research, AI, security/privacy, interoperability, blockchain, fault management, or applied Metaverse evidence are absent. The defensible residual gaps are narrower: cross-paper comparability, artifact provenance and independent reproduction, standards conformance and adoption, native immersive-security transfer, direct sustainability measurement, and cross-layer experimental evidence. The result is an auditable survey architecture that separates what the literature contains from what its evidence can actually support.