Reproduce “NetIntent: Leveraging Large Language Models for End-to-End Intent-Based SDN Automation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Md. Kamrul Hossain, Walid Aljoby

Researchers at King Fahd University of Petroleum & Minerals systematically benchmarked 33 open-source Large Language Models for Intent-Based Networking tasks and developed NetIntent, an end-to-end framework automating the full intent lifecycle in Software-Defined Networks. The framework successfully translates, activates, and assures natural language intents on OpenDaylight and ONOS controllers, achieving end-to-end realization times between approximately 17.90s and 35.10s.

Key insights

  • LLM performance for IBN tasks is highly dependent on prompt design, schema alignment, and in-context learning, with specialized or well-prompted mid-sized models often outperforming larger, general-purpose models at a lower computational cost.
  • Accurate conflict detection between network policies using LLMs requires precise structural reasoning and literal matching rules, which only a few top-performing models (e.g., QwQ:32b) consistently achieve without high false positive rates.
  • An end-to-end framework combining LLM-driven natural language processing with non-LLM agents for validation, deterministic conflict resolution, and continuous assurance provides a robust approach to automate the full IBN lifecycle.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.