Codebase generation on Event Planner App
62Feature Completion (%)Code-L2MAC
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Code-L2MAC2023.10 | 62 | 0 | 473 | 25.6 | 97.1 | |
| CodeT2023.10 | 26.2 | 0.05 | 75.2 | 2.45 | 92.5 | |
| Reflexion2023.10 | 23.6 | 0 | 82 | 3 | 95.1 | |
| AutoGPT2023.10 | 22.5 | 0 | 23.9 | 0 | 32.9 | |
| Self-Refine2023.10 | 21.4 | 0.15 | 118 | 3.9 | 76.7 | |
| GPT42023.10 | 11.2 | 0.025 | 74.6 | 1.75 | 88.7 |