| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| WG (test) | Qwen2.5-14B-IT + SELF-REDTEAM | ASR0 | 36 | 18d ago | |
| DAN | Qwen2.5-3B-IT | ASR94.9 | 26 | 18d ago | |
| WJB | SELF-REDTEAM | ASR0.2 | 26 | 18d ago | |
| StrongREJECT vani harm | Qwen2.5-14B-IT + SELF-REDTEAM | RTA Rate (vani harm)99.5 | 10 | 18d ago | |
| XSTest vani harm | Llama3.1-8B-IT + SELF-REDTEAM | RTA Score94.3 | 10 | 18d ago | |
| OR-Bench vani harm | Qwen2.5-14B-IT + SELF-REDTEAM | RTA98.5 | 10 | 18d ago | |
| HarmBench vani harm | Qwen2.5-14B-IT + SELF-REDTEAM | ASR1 | 10 | 18d ago | |
| HarmBench adv harm | Qwen2.5-14B-IT + SELF-REDTEAM | ASR6.4 | 10 | 18d ago | |
| StrongREJECT Adversarial | SELF-REDTEAM | Attack Success Rate (ASR)37.1 | 9 | 2mo ago | |
| XSTest Adversarial | SELF-REDTEAM | RTA92.5 | 9 | 2mo ago | |
| Sorry-Bench | LlamaGuard prompt-filter | Refusal Rate (String Match)85.5 | 6 | 1mo ago | |
| AdvBench | LlamaGuard prompt-filter | Refusal Rate (String Match)99.4 | 6 | 1mo ago | |
| HarmBench Contextual | LlamaGuard response-filter | String-match Refusal Rate100 | 6 | 1mo ago | |
| HarmBench Standard | LlamaGuard prompt-filter | Refusal Rate (String Match)98.2 | 6 | 1mo ago |
