Red-teaming on CatQA
0ASRSafeTransformer
Evaluation results
| Method | Links | |
|---|---|---|
| SafeTransformerPrompt=Standard, Inference Mode=eos2026.03 | 0 | |
| SafeTransformerPrompt=CoT, Inference Mode=eos2026.03 | 0 | |
| SafeTransformerPrompt=CoU, Inference Mode=eos2026.03 | 0 | |
| SafeTransformerPrompt=Standard, Inference Mode=avg2026.03 | 0 | |
| SafeTransformerPrompt=CoT, Inference Mode=avg2026.03 | 0 | |
| SafeTransformerPrompt=CoU, Inference Mode=avg2026.03 | 0 | |
| SafeTransformerPrompt=Suffix, Inference Mode=avg2026.03 | 0 | |
| SafeTransformerPrompt=Avg., Inference Mode=avg2026.03 | 0 | |
| SafeTransformerPrompt=Avg., Inference Mode=eos2026.03 | 0.18 | |
| SafeTransformerPrompt=Suffix, Inference Mode=eos2026.03 | 0.73 | |
| Llama-3.2-1BPrompt=CoU2026.03 | 11.09 | |
| Llama-3.2-1BPrompt=Suffix2026.03 | 16.18 | |
| Llama-3.2-1B (SFT)Prompt=Suffix2026.03 | 16.18 | |
| Llama-3.2-1B (SFT)Prompt=Standard2026.03 | 18.73 | |
| Llama-3.2-1B (SFT)Prompt=Avg.2026.03 | 25.46 | |
| Llama-3.2-1BPrompt=Avg.2026.03 | 28.09 | |
| Llama-3.2-1B (SFT)Prompt=CoT2026.03 | 30.55 | |
| Llama-3.2-1BPrompt=Standard2026.03 | 34.91 | |
| Llama-3.2-1B (SFT)Prompt=CoU2026.03 | 36.36 | |
| Llama-3.2-1BPrompt=CoT2026.03 | 50.18 |