Hyperbolic embeddings for graph compression

Feed of July 14, 2026 arXiv CS items covering graph methods, knowledge graphs, LLMs, and social-media studies. Notable security-relevant papers: “Large Language Models in Misinformation Ecosystems” (systematizes LLM-enabled attacks on content, evidence sources, retrieval corpora, and verification workflows and calls out vulnerabilities and defenses); “MafiaScope” (provides a measurement/testbed for probing internal beliefs of LLM agents — potential for privacy/leakage and model interrogation); work showing LLMs can reproduce human decision biases (implications for scalable behavioral modeling,