Tag
The paper evaluates a deployed multi-agent system for formal tender responses, demonstrating it matches human quality in evaluations and highlights the asymmetry where structural markup aids reading but not writing.