| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Jailbreak Attack | Claude 3.5 | ASR0 | 24 | |
| Jailbreak Attack | Claude Sonnet API 3.5 | ASR80.5 | 16 | |
| AI-Generated Text Detection | Claude Sonnet 3.7 | AUROC (Insertion)0.9964 | 10 | |
| Black-box Adversarial Attack | Claude thinking 4.0 | KMR (a)0.02 | 9 | |
| Jailbreaking | Claude 4.5 | ASR97 | 9 | |
| Targeted Attack | Claude-3-Opus 4.6 (test) | ASR76.8 | 8 | |
| Targeted Adversarial Attack | Claude 4.7 | ASR76.5 | 8 | |
| Targeted Adversarial Attack | Claude 4.6 | Attack Success Rate (ASR)69.2 | 8 | |
| AI-generated text detection | Claude-generated (test) | F1 Score92.2 | 5 | |
| Keyword Matching Attack | Claude-3-Opus | KMR (alpha)92 | 4 | |
| Adversarial Attack Transfer | Claude 3.5 | Similarity Score (SS)64.5 | 3 |