Benchmarks Reproducing

Reproducing Benchmark Results

Commands to reproduce all three benchmark tiers locally:

---

1. 8-Way Wide Fanout Swarm Benchmark ($N=50$)

cd Phase1
PYTHONPATH=python_adapter uv run python benchmarks/benchmark_wide_fanout_swarm.py --trials 50

---

2. Trace-Replayed Sequential Reflection Swarm ($N=50$)

cd Phase1
PYTHONPATH=python_adapter uv run python benchmarks/benchmark_replayed_swarm.py --trials 50

---

3. Live Cloud LLM Financial Swarm (Groq LPU)

export GROQ_API_KEY=your_key_here
cd Phase1
PYTHONPATH=python_adapter uv run python python_adapter/financial_swarm.py AAPL
```\n