Agentic Interpretation: Lattice-Structured Evidence for LLM-Based Program Analysis

arXiv:2605.12694v1 Announce Type: cross Abstract: Large language models can consult information that fixed static analyzers cannot, such as documentation, current security advisories, version-specific metadata, and informal API contracts. This makes LLMs a compelling option for program analyses...