Tag
PlayCoder introduces PlayEval benchmark and a multi-agent framework that iteratively repairs LLM-generated GUI applications, achieving up to 20.3% end-to-end playable code.