List your company

Over-Safety Evaluation on XSTest (XST Metric)

23.2XST Over-SafetyTELLME
5.72810.26414.819.336Feb 7, 2025

Evaluation results

MethodLinks
TELLMEBackbone=Qwen2.5-7B2025.0223.2
TELLME NT-XentBackbone=Qwen2.5-7B2025.0222.4
TELLME NT-XentBackbone=Llama-3.1-8B2025.0221.2
TELLMEBackbone=Llama-3.1-8B2025.0218
SFTBackbone=Llama-3.1-8B2025.0216.4
OriginBackbone=Qwen2.5-7B2025.0216
SFTBackbone=Mistral-7B-v0.32025.0215.6
OriginBackbone=Mistral-7B-v0.32025.0214.4
SFTBackbone=Qwen2.5-7B2025.0212
TELLME NT-XentBackbone=Mistral-7B-v0.32025.0210.8
TELLMEBackbone=Mistral-7B-v0.32025.029.2
OriginBackbone=Llama-3.1-8B2025.026.4