FAVA: Formal Authorization for Verified Agents with Evidence-Backed Permission Graphs

Yifan Zhang, Xinkui Zhao, Sai Liu +3

Large language model (LLM) agents autonomously interleave semantic reasoning with complex system operations. In these dynamic environments, static tool-level permissions are fundam…