List your company

Safety Evaluation on WildJailbreak (held-out)

0Attack Success Rate (ASR)NeST
-3.16418.19339.5560.907Feb 18, 2026

Evaluation results

MethodLinks
NeSTTarget Model=Phi-42026.020
Full FTTarget Model=gemma-2b-it2026.020.1
Full FTTarget Model=Phi-4-mini2026.020.3
Full FTTarget Model=gemma-7b-it2026.020.4
Full FTTarget Model=Qwen3-14B2026.020.6
NeSTTarget Model=gemma-2b-it2026.020.7
Full FTTarget Model=Qwen3-4B2026.020.7
Full FTTarget Model=Qwen2.5-14B2026.020.8
LoRATarget Model=Llama-3.2-1B2026.021.4
NeSTTarget Model=Qwen3-4B2026.021.5
Full FTTarget Model=Qwen2.5-7B2026.021.6
LoRATarget Model=Qwen3-4B2026.021.6
NeSTTarget Model=Llama-3.2-1B2026.021.7
Full FTTarget Model=Llama-3.2-3B2026.021.9
Full FTTarget Model=Phi-42026.021.9
Full FTTarget Model=Llama-3.2-1B2026.022.1
NeSTTarget Model=Llama-3.2-3B2026.022.8
LoRATarget Model=Qwen3-14B2026.022.8
NeSTTarget Model=Qwen2.5-14B2026.022.9
LoRATarget Model=Llama-3.2-3B2026.023.4
NeSTTarget Model=gemma-7b-it2026.024.1
LoRATarget Model=gemma-2b-it2026.024.2
LoRATarget Model=Phi-4-mini2026.025.7
LoRATarget Model=gemma-7b-it2026.026.6
NeSTTarget Model=Phi-4-mini2026.026.7
NeSTTarget Model=Qwen2.5-7B2026.0211
NeSTTarget Model=Qwen3-14B2026.0212.2
LoRATarget Model=Qwen2.5-14B2026.0213
LoRATarget Model=Phi-42026.0217.8
LoRATarget Model=Qwen2.5-7B2026.0219.2
BaselineTarget Model=gemma-2b-it2026.0222
BaselineTarget Model=Phi-42026.0225.2
BaselineTarget Model=Phi-4-mini2026.0225.7
BaselineTarget Model=gemma-7b-it2026.0234
Circuit BreakerTarget Model=gemma-2b-it, Protocol=Exp.2026.0239.7
BaselineTarget Model=Llama-3.2-1B2026.0240.5
Circuit BreakerTarget Model=gemma-7b-it, Protocol=Exp.2026.0240.6
Circuit BreakerTarget Model=Phi-4-mini, Protocol=Exp.2026.0244.5
BaselineTarget Model=Qwen3-14B2026.0250.9
BaselineTarget Model=Llama-3.2-3B2026.0254.2
Circuit BreakerTarget Model=Llama-3.2-1B, Protocol=Exp.2026.0258.1
BaselineTarget Model=Qwen2.5-14B2026.0258.1
BaselineTarget Model=Qwen3-4B2026.0265.1
Circuit BreakerTarget Model=Qwen2.5-14B, Protocol=Exp.2026.0265.2
Circuit BreakerTarget Model=Phi-4, Protocol=Exp.2026.0268.2
BaselineTarget Model=Qwen2.5-7B2026.0269.4
Circuit BreakerTarget Model=Qwen3-14B, Protocol=Exp.2026.0272.1
Circuit BreakerTarget Model=Llama-3.2-3B, Protocol=Exp.2026.0274.8
Circuit BreakerTarget Model=Qwen3-4B, Protocol=Exp.2026.0276.4
Circuit BreakerTarget Model=Qwen2.5-7B, Protocol=Exp.2026.0279.1