Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Evasion Detection on CLI positional attacks
Loading...
31.5
Evasion Rate
4-monitor ensemble
28.832
46.841
64.85
82.859
Jul 2, 2026
Evasion Rate
Updated 22d ago
Evaluation Results
Method
Method
Links
Evasion Rate
4-monitor ensemble
Threshold=mean > 3.75/...
2026.07
31.5
Diff: link-tracker
Threshold=6
2026.07
62.1
Traj: full-trajectories
Threshold=3
2026.07
65.1
Traj: task-description
Threshold=3
2026.07
71.9
Diff: task-description
Threshold=6
2026.07
75.8
Diff: cumulative
Threshold=5
2026.07
75.8
Diff: full-diffs
Threshold=7
2026.07
98.2
Feedback
Search any
task
Search any
task