Agentic Interpretation: Lattice-Structured Evidence for LLM-Based Program Analysis
For program analysis tasks requiring information beyond source code, this framework provides a principled way to leverage LLMs while maintaining the rigor of static analysis, addressing brittleness of one-shot LLM analyses.
The paper proposes agentic interpretation, a framework combining lattice-based static analysis with LLM-driven program reasoning to decompose analysis goals into localized claims tracked in a lattice, enabling more robust and evidence-aware program analysis. The approach is illustrated with a worked example analyzing code dependent on opaque third-party components.
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 that depend on information beyond the source program, or that are otherwise not amenable to conventional static analyzers. However, directly asking an LLM for a one-shot whole-program analysis is brittle because it compresses many evidence-dependent judgments into a single opaque answer, rather than exposing which conclusions are supported or disputed and using intermediate findings to guide later, more focused searches. In this paper, we propose agentic interpretation, a framework that brings the discipline of lattice-based static analysis to LLM-driven program reasoning. At a high level, agentic interpretation decomposes a high-level analysis goal into localized claims, and tracks the LLM's judgment about each claim in a finite-height lattice. A worklist algorithm governs how claims and their judgments evolve during the analysis. We introduce a formal model of agentic interpretation, explore the design space it opens, and illustrate the approach with a worked example analyzing code that depends on opaque third-party components.