List your company

Safety Evaluation on WildTeaming 500-example (test)

88.6HarmRFt. Agent
53.55262.65171.7580.849Oct 19, 2025

Evaluation results

MethodLinks
Ft. AgentBackbone=Qwen-2.5-3B-Instruct2025.1088.62.87
Naive RAGBackbone=Qwen-2.5-3B-Instruct2025.1088.52.84
Base AgentBackbone=Qwen-2.5-3B-Instruct2025.1087.52.9
Naive RAGBackbone=Qwen-2.5-7B-Instruct2025.1087.32.97
Ft. AgentBackbone=Qwen-2.5-7B-Instruct2025.1087.32.88
Base LLMBackbone=Qwen-2.5-3B-Instruct2025.10872.84
Base AgentBackbone=Qwen-2.5-7B-Instruct2025.1083.92.79
Base LLMBackbone=Qwen-2.5-7B-Instruct2025.1081.72.73
Naive RAGBackbone=Qwen-2.5-14B-Instruct2025.1077.42.45
Base AgentBackbone=Qwen-2.5-14B-Instruct2025.1069.12.43
Base LLMBackbone=Qwen-2.5-14B-Instruct2025.1068.52.4
Base LLMBackbone=Mistral-NeMo-8B-Instruct2025.1056.33.29
Naive RAGBackbone=Mistral-NeMo-8B-Instruct2025.1054.93.17
Base AgentBackbone=Mistral-NeMo-8B-Instruct2025.1054.93.17