When Code Becomes an LLM Call: An Empirical Study of Call-Site Contracts in Agentic Systems

Key points

• Software quality has long rested on specifications that state what code must do, and on testing and verification that check that the code does it.1
• Agentic systems change this foundation: they replace parts of their logic, such as choosing the next tool, judging whether an action is allowed, or writing code to run, with calls to a large language model (LLM).1
• Using a multi-agent pipeline validated by 17,428 human judgments, we trace where every model output flows, derive an expected contract for each call site without access to its code, and compare it with the contract that the code implements.1
• Our results show that the presence of checks says little about their adequacy, and we derive implications for agent developers, framework maintainers, and researchers.1

Sources (1)

• [1]When Code Becomes an LLM Call: An Empirical Study of Call-Site Contracts in Agentic Systems
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 17, 08:14 AM

  We present an empirical study of these contracts in 496 LLM call sites of 40 popular open-source agentic systems (13.4 MLOC, four languages).

  Software quality has long rested on specifications that state what code must do, and on testing and verification that check that the code does it.

Extractive summary: sentences quoted from the sources.