Adversarial Robustness (GCG attack) on StrongREJECT (sample of 40 prompts)
26.87Mean StrongREJECT ScoreLlama 2 7B Chat
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama 2 7B ChatModel=Llama 2 7B Chat, Fine-tuning=None2025.12 | 26.87 | |
| Llama 2 7B Chat (DA)Model=Llama 2 7B Chat, Fine-tuning=Data Augmentation (DA)2025.12 | 10.94 | |
| Llama 2 7B Chat (DA+PRESTO)Model=Llama 2 7B Chat, Fine-tuning=Data Augmentation (DA) + PRESTO2025.12 | 5.63 |