empirical-systems-evaluation

Benchmark multi-agent coordination systems with experiment design, power analysis, human rating protocols, bootstrap confidence intervals, and reproducible reporting. Use for salvage latency, recovery fidelity, coordination overhead, and protocol comparison studies. NOT for ML model benchmarking, web-product A/B testing, survey design, or general-purpose data science.