Multilateral Clearing on Invoice Graphs: Path Enabled Compensation Versus Cycle Restricted Netting

Collection of arXiv CS submissions (26 Jun 2026) covering network science, graph algorithms, privacy, and adversarial ML. Key papers: (1) “Multilateral Clearing on Invoice Graphs” introduces Byppay, a path-enabled local compensation procedure that substantially increases debt relief over cycle-restricted netting (empirical evaluation on a large invoice corpus); (2) “Account-History Features for Social Bot Detection” shows account-history features (age, follower/friend ratios, profile structure) remain highly predictive (AUC≈0.977) even when adversaries rewrite text using LLMs, arguing content-