Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Prompt Injection Detection on SCOUT-450 (Held-out evaluation)
Loading...
92.4
Accuracy
GPT-4o
65.88
72.765
79.65
86.535
May 29, 2026
Accuracy
Attack Success Rate
False Positive Rate
Updated 2d ago
Evaluation Results
Method
Method
Links
Accuracy
Attack Success Rate
False Positive Rate
GPT-4o
Lat (ms)=1457
2026.05
92.4
11.8
2.1
GPT-5.1
Lat (ms)=1464
2026.05
87.3
16.9
7.2
DeepSeek-V4
Lat (ms)=9224
2026.05
87.1
21.6
1.5
GPT-5.2
Lat (ms)=3638
2026.05
80.4
5.1
38.5
Gemini-3.1
Lat (ms)=4806
2026.05
66.9
55.7
3.6
Feedback
Search any
task
Search any
task