Reproduce “Automatically Generating Web Applications from Requirements Via Multi-Agent Test-Driven Development” 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
Yuxuan Wan, Tingshuo Liang, Jiakai Xu, Jingyu Xiao, Yintong Huo, Michael R. Lyu
TDDev is presented as a multi-agent, Test-Driven Development (TDD) enabled framework that automatically generates end-to-end full-stack web applications directly from user requirements. This system demonstrates an overall accuracy of 78.2% with GPT-4.1, representing a 30% improvement over state-of-the-art baselines like Cursor, and achieves a 0% application failure-to-start rate.
Key insights
- A multi-step test case generation process (decomposition, elaboration, soap-opera testing) significantly improves functionality and overall accuracy compared to single-shot test generation.
- Iterative feedback from the Testing Agent, particularly after three rounds, is crucial for refining both functional correctness and visual fidelity of the generated applications.
- The multi-agent TDD approach allows for the elimination of manual developer intervention during the application generation process, enhancing efficiency despite a potentially longer total generation time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.