# Reproduction: FormulaCode: Evaluating Agentic Optimization on Large Codebases

## Pages

| Page |
| --- |
| [Executive summary](#/executive-summary) |
| [Claim 1: Dataset construction pipeline](#/claim-1-dataset-construction-pipeline) |
| [Claim 2: Workloads and patch size](#/claim-2-workloads-and-patch-size) |
| [Claim 3: Main agent results](#/claim-3-main-agent-results) |
| [Claim 4: Optimization scale](#/claim-4-optimization-scale) |
| [Claim 5: Repository popularity effects](#/claim-5-repository-popularity-effects) |
| [Claim 6: Optimization type performance](#/claim-6-optimization-type-performance) |
| [Fresh independent CPU audit](#/claim-99-fresh-independent-cpu-audit) |
| [Conclusion](#/conclusion) |