Coding Performance on Insecure-code 1000-prompt held-out
93.9Task Success Rate
Naive (corruption baseline)
Updated 6d ago
Evaluation Results
| Method | Links | |
|---|---|---|
2026.05 | 93.9 | |
2026.05 | 93.6 | |
2026.05 | 92.3 | |
2026.05 | 87.3 | |
2026.05 | 86.2 | |
2026.05 | 84.8 | |
2026.05 | 78.8 |
Naive (corruption baseline)
| Method | Links | |
|---|---|---|
2026.05 | 93.9 | |
2026.05 | 93.6 | |
2026.05 | 92.3 | |
2026.05 | 87.3 | |
2026.05 | 86.2 | |
2026.05 | 84.8 | |
2026.05 | 78.8 |