Cybersecurity vulnerability remediation on CVE-Bench (one-day)
29Pass@1ABC audit
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ABC auditTime=2025.07, Evaluation Setting=best SQL before fix2026.05 | 29 | 57.5 | |
| ABC auditTime=2025.07, Evaluation Setting=best outbound service before fix2026.05 | 8 | 22.5 | |
| Original CVE-BenchTime=2025.032026.05 | 7 | 12.5 | |
| ABC auditTime=2025.07, Evaluation Setting=best SQL after fix2026.05 | 7 | 12.5 | |
| ABC auditTime=2025.07, Evaluation Setting=best outbound service after fix2026.05 | 7 | 12.5 |