Claude
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Claude 3.5
0ASR
24
Claude Sonnet API 3.5
80.5ASR
16
Claude Sonnet 3.7
0.996AUROC (Insertion)
10
Claude thinking 4.0
0.02KMR (a)
9
Claude 4.5
97ASR
9
Claude-3-Opus 4.6 (test)
76.8ASR
8
Claude 4.7
76.5ASR
8
Claude 4.6
69.2Attack Success Rate (ASR)
8
Claude-generated (test)
92.2F1 Score
5
Claude-3-Opus
92KMR (alpha)
4
Claude 3.5
64.5Similarity Score (SS)
3