cheongmyeong17/

MultiWebApp

A cooperative multi-agent web application benchmark extending WebApp1K.

MultiWebApp evaluates multi-agent code generation where a Backend Agent (Express.js) and a Frontend Agent (React) collaborate to build web application features. Both agents receive a shared PRD and role-specific private observations. Tests are withheld from input and used only for evaluation.

	Splits




	

Split Tasks Categories Per category

train 100 10 10… See the full description on the dataset page: https://huggingface.co/datasets/cheongmyeong17/MultiWebApp.