Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Attack Detection on Agent Traces (Held-out)
Loading...
0.89
ROC-AUC (Social engineering)
gated multi-view fusion
0.6612
0.7206
0.78
0.8394
Jan 5, 2026
ROC-AUC (Social engineering)
ROC-AUC (Prompt injection)
ROC-AUC (Data exfiltration)
ROC-AUC (Tool hijacking)
ROC-AUC (Unknown)
Updated 4d ago
Evaluation Results
Method
Method
Links
ROC-AUC (Social engineering)
ROC-AUC (Prompt injection)
ROC-AUC (Data exfiltration)
ROC-AUC (Tool hijacking)
ROC-AUC (Unknown)
gated multi-view fusion
Representation=Gated
2026.01
0.89
0.83
0.62
0.6
0.92
Conversational tokenization
Representation=Convers...
2026.01
0.78
0.69
0.46
0.39
0.26
structural tokenization
Representation=Structural
2026.01
0.67
0.81
0.85
0.85
0.97
Feedback
Search any
task
Search any
task