Vulnerability Detection on CyberGym 100 projects
175Vulnerabilities Detected CountREVELIO
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| REVELIOModel=Claude Haiku 4.5 → Claude Sonnet 4.62026.06 | 175 | 9.1 | 69 | 0 | 175 | 19,208,917 | |
| Claude CodeModel=Claude Opus 4.7*2026.06 | 55 | 16.2 | 53 | 49 | 107 | 17,958,495 | |
| CodexModel=GPT 5.52026.06 | 39 | 8.96 | 39 | 60 | 97 | 10,612,921 | |
| SorcarModel=Claude Opus 4.72026.06 | 31 | 6.27 | 31 | 28 | 43 | 5,042,576 |