Cybersecurity Challenge Solving on Cybench 33-challenge
15Solved CountCSI::Claude
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| CSI::ClaudeScaffold=Claude2026.05 | 15 | 45.5 | 26.8 | 5,122 | 341 | 8,370 | 2,554 | 0.001 | 14.6 | |
| CSI::CodexScaffold=Codex2026.05 | 15 | 45.5 | 18.4 | 1,713 | 114 | 3,437 | 287 | 0.339 | 3.9 | |
| CSI::GCAIScaffold=GCAI2026.05 | 10 | 30.3 | 30.4 | 1,279 | 128 | 9,734 | 809 | 0.294 | 10.3 | |
| CSI::CAIScaffold=CAI2026.05 | 7 | 21.2 | 15.9 | 727 | 104 | 386 | 454 | 0.159 | 1.1 |