Show HN: HATS — Hierarchical Agent Teams That Debate to Improve Decision Quality
GitHub / Hacker News (28pts, 17 comments)·low signal
Open-source framework implementing multi-agent debate where AI agents with distinct roles argue and reach consensus before producing output. Based on research showing adversarial multi-agent debate yields 4-6% higher accuracy and 30%+ fewer factual errors compared to single-agent methods on mathematical reasoning and factual tasks. Practical implementation of the debate-as-quality-improvement pattern that's gaining traction as teams scale from single-agent to multi-agent production architectures.