Automated Code and Document Review on 30 AI-generated artifacts with 150 injected errors
5.5Total FindsSelf-Review ×2 (SR2)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Self-Review ×2 (SR2)Model=Claude Opus 4.6, Context=Full + 1st review2026.03 | 5.5 | 1.1 | 4.4 | 3.9 | 21 | 22.7 | 21.7 | |
| Self-Review (SR)Model=Claude Opus 4.6, Context=Full history2026.03 | 4.8 | 1.2 | 3.6 | 3.8 | 25.8 | 24.2 | 24.6 | |
| Cross-Context Review (CCR)Model=Claude Opus 4.6, Context=Artifact only2026.03 | 4.5 | 1.4 | 3.1 | 3.6 | 31.5 | 27.1 | 28.6 | |
| Subagent Review (SA)Model=Claude Opus 4.6, Context=Prompt + artifact2026.03 | 4.3 | 1.1 | 3.2 | 3.9 | 27.4 | 21.8 | 23.8 |