Safety Classification on Sorry-Bench (test)
93.9Accepted HF AccuracyLatent Guard
Evaluation Results
| Method | Links | |
|---|---|---|
| Latent GuardBase Model=LLAMA2-CHAT-7B2025.07 | 93.9 | |
| Llama Guard 3Base Model=QWEN2-INSTRUCT-7B2025.07 | 59.4 | |
| Latent GuardBase Model=LLAMA3-INSTRUCT-8B2025.07 | 59.3 | |
| Latent GuardBase Model=QWEN2-INSTRUCT-7B2025.07 | 54.6 | |
| Llama Guard 3Base Model=LLAMA2-CHAT-7B2025.07 | 45.5 | |
| Llama Guard 3Base Model=LLAMA3-INSTRUCT-8B2025.07 | 37.3 |