Safety Evaluation on BeaverTails (BT)
99.1BT ScoreTELLME NT-Xent
Evaluation Results
| Method | Links | |
|---|---|---|
| TELLME NT-XentBackbone=Qwen2.5-7B2025.02 | 99.1 | |
| TELLME NT-XentBackbone=Mistral-7B-v0.32025.02 | 99 | |
| TELLMEBackbone=Qwen2.5-7B2025.02 | 98.7 | |
| TELLME NT-XentBackbone=Llama-3.1-8B2025.02 | 97.1 | |
| TELLMEBackbone=Mistral-7B-v0.32025.02 | 96.2 | |
| TELLMEBackbone=Llama-3.1-8B2025.02 | 95.5 | |
| SFTBackbone=Llama-3.1-8B2025.02 | 95 | |
| SFTBackbone=Mistral-7B-v0.32025.02 | 93.7 | |
| OriginBackbone=Qwen2.5-7B2025.02 | 92.1 | |
| OriginBackbone=Mistral-7B-v0.32025.02 | 84.3 | |
| OriginBackbone=Llama-3.1-8B2025.02 | 83.1 | |
| SFTBackbone=Qwen2.5-7B2025.02 | 58.4 |