Prompt Injection Attack Evaluation on WalledEval
100ASPNeural-chat
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Neural-chatOpen-source=true2025.05 | 100 | 100 | |
| Stablelm2Open-source=true2025.05 | 97.3 | 97.3 | |
| MistralOpen-source=true2025.05 | 87.3 | 73.3 | |
| OpenchatOpen-source=true2025.05 | 78.7 | 63.3 | |
| Starling-lmOpen-source=true2025.05 | 75.7 | 62 | |
| VicunaOpen-source=true2025.05 | 73.3 | 71.3 | |
| PhiOpen-source=true2025.05 | 66.7 | 57.3 | |
| Gemma2Open-source=true2025.05 | 65.3 | 23.3 | |
| Deepseek-r1Open-source=true2025.05 | 54.3 | 47.3 | |
| Llama2Open-source=true2025.05 | 50 | 0 | |
| Phi3Open-source=true2025.05 | 25.3 | 0.7 | |
| Llama3Open-source=true2025.05 | 6 | 0.7 | |
| GemmaOpen-source=true2025.05 | 5 | 3.3 | |
| Gemma-2bOpen-source=true2025.05 | 2.3 | 1.3 |