Reachability Across the NL/PL Boundary: A Taxonomy-Driven Dataflow Model for LLM-Integrated Applications
arXiv:2603.28345v3 Announce Type: replace-cross Abstract: LLM API calls have become a standard programming primitive, but they create a program boundary that disrupts traditional dataflow analysis. A runtime value may be inserted into a natural-language prompt through a template placeholder, transformed opaquely by the LLM, and returned as code, JSON, or text consumed by downstream logic. Existing analyses such as taint analysis and program slicing require a dataflow summary that describes how a callee maps inputs to outputs; an LLM call provides no such summary, breaking analysis at what we c
The proliferation of LLM API calls as a programming primitive necessitates new methods for understanding and analyzing dataflow within these hybrid natural language/programming language systems.
This research addresses a fundamental challenge in software engineering for LLM-integrated applications, impacting reliability, security, and debugging of agentic systems.
Traditional software analysis tools are inadequate for LLM-integrated applications, requiring a new taxonomy and dataflow model to describe how LLMs transform data opaquely.
- · Software developers
- · Cybersecurity firms
- · AI platform providers
- · Legacy software analysis tools
- · Organizations relying solely on traditional development methods
Improved reliability and security of software incorporating large language models become possible.
Faster development and deployment of complex AI agentic systems due to better debugging and understanding of data flow.
New programming paradigms and tools emerge that are specifically designed for the unique challenges of NL/PL boundaries in software.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI