Jailbreaking on Various SOTA Jailbreaking Benchmarks (test)
92Best ASRAutoDAN
Evaluation Results
| Method | Links | |
|---|---|---|
| AutoDANYear=2023, Target=Llama-2, Key Approach=Automated prompt generation2026.06 | 92 | |
| GCGYear=2023, Target=GPT-2, Key Approach=Gradient-based optimization2026.06 | 84 | |
| TAPYear=2023, Target=GPT-4, Key Approach=Tree-based search2026.06 | 84 | |
| JailbreakBenchYear=2024, Target=Multiple, Key Approach=Standardized benchmark2026.06 | 78 | |
| PAIRYear=2023, Target=GPT-4, Key Approach=Iterative refinement2026.06 | 60 | |
| Red Teaming Framework for LLMsYear=2025, Target=Multiple, Key Approach=Cross-linguistic evaluation2026.06 | 0 |