2 papers
cs.MA2026
Where Facts Go Missing: A Layerwise Taxonomy and Per-Layer Attribution of Information Omission in Air-Gapped LLMAgent Pipelines
Santhiya Rajan, Samuel Mugel, Roman Orus
Air-gapped and on-premises language-model agents can silently omit decision-critical facts at any boundary between source ingestion and final answer generation. We present a nine-l…
cs.AI2026
ProvenanceGuard: Source-Aware Factuality Verification for MCP-Based LLM Agents
Ander Alvarez, Santhiya Rajan, Alessandro Genuardi +3
Tool-using LLM agents increasingly use the Model Context Protocol (MCP) to answer from heterogeneous evidence sources, including search, APIs, databases, clinical records, and form…