Jailbreak Defense on Emoji Attack MultiJail
0Attack Success Rate (ASR)SelfGrader
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SelfGraderTarget LLM=LLama-3-8B Instruct2026.04 | 0 | 0 | |
| Llama Guardvariant=Pre, Target LLM=LLama-3-8B Instruct2026.04 | 6.9 | 100 | |
| SelfDefendvariant=Direct, Target LLM=LLama-3-8B Instruct2026.04 | 6.98 | 89.2 |