AGT-09

Multi-Agent Debate Evaluation

Compare independent, debate, critic, and ensemble patterns on a labeled benchmark; report when extra agents hurt cost or quality.

• Status: planned
• Proof: code
• Tier: focused-exhibit
• Roles: ai-engineer, backend-engineer

← All projects