Reproduce “Agent4RE: A Self-Refining Multi-agent Framework for End-to-End Software Requirements Engineering and Benchmarking” 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

Yongjian Tang, Linhan Li, Thomas Runkler

Agent4RE uses specialist agents to elicit project needs, draft a requirements specification, and revise it with agent or human feedback. The authors evaluate it on RE-E2E, a benchmark reported to contain 50 specifications. Across eight LLMs, structured human feedback has a mean human rating of 2.74 out of 4, versus 1.80 for a prompting-pattern baseline; this benchmark result does not establish industrial benefit.

Summary by alphaXiv.

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