Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Agent Safety Judgment on ROME IR unsafe
Loading...
95.74
F1 Score
Claude Sonnet 4.6
82.532
85.961
89.39
92.819
May 5, 2026
F1 Score
Recall
Validity
Updated 28d ago
Evaluation Results
Method
Method
Links
F1 Score
Recall
Validity
Claude Sonnet 4.6
Model=Claude Sonnet 4.6
2026.05
95.74
91.84
98
GPT-5.2
Model=GPT-5.2
2026.05
91.3
84
100
Claude Opus 4.6
Model=Claude Opus 4.6
2026.05
86.06
75.53
94
GPT-5
Model=GPT-5
2026.05
83.04
71
100
Feedback
Search any
task
Search any
task