Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Vulnerability Discovery on 847 historical Linux kernel CVEs Replay Mode 2019-2025
Loading...
3.2
Time To First Vulnerability (hrs)
VCAO (full)
2.76
5.73
8.7
11.67
Apr 9, 2026
Time To First Vulnerability (hrs)
SVUB
FPR (%)
Sibling Count/Score
Precision (%)
Updated 9d ago
Evaluation Results
Method
Method
Links
Time To First Vulnerability (hrs)
SVUB
FPR (%)
Sibling Count/Score
Precision (%)
VCAO (full)
Sibling search=true
2026.04
3.2
1.13
15.1
2.4
67.8
VCAO (no sib.)
Sibling search=false (...
2026.04
3.8
1.02
15.3
0
61.7
Multi-agent (no GT)
Ranking=LRM, Stackelbe...
2026.04
5.1
0.81
24.6
1.3
48.5
Fuzz-only
Engine=Syzkaller, Conf...
2026.04
8.4
0.42
31.4
0
26.7
Static-only
Engine=CodeQL, Configu...
2026.04
9.7
0.59
47.3
0
31.2
Churn-based
Ranking strategy=git c...
2026.04
11.8
0.39
38.2
0
22.1
Uniform
Allocation strategy=Eq...
2026.04
14.2
0.31
42.7
0
18.3
Feedback
Search any
task
Search any
task