List your company

Jailbreak Attack Success on SpatialJB Summary 1.0

76Acrostic ASRLlama4
41.47250.43659.468.364Jan 14, 2026

Evaluation results

MethodLinks
Llama4Guardrail=None2026.017673.775.375.674.376.39.78.49.5
Grok4Guardrail=None2026.017575757575759.58.28.9
Deepseek-R1Guardrail=None2026.017474.574.273.77575.38.26.27.6
Deepseek-V3Guardrail=None2026.0173.775.374.674.673.875.38.76.47.8
GPT4Guardrail=None2026.0171.271.772.870.77373.77.84.87
Gemini 2.5-ProGuardrail=None2026.0170.368.567.369.869.870.76.44.57.9
Deepseek-V3Guardrail=OpenAI2026.0168.866.568.858.76967.56.24.55.8
Deepseek-V3Guardrail=Perspective2026.0168.569.26867.368.266.265.85.3
GPT4Guardrail=Llama Guard2026.0168.267.368.265.86869.25.33.95.1
Deepseek-V3Guardrail=Llama Guard2026.0168.268.268.264.569.865.25.666.6
GPT4Guardrail=OpenAI2026.01686567.762.268.271.25.12.95.8
Llama4Guardrail=Perspective2026.0168676766.568.567.26.65.34.5
Grok4Guardrail=Perspective2026.016862.969.265.869.864.98.26.47.6
Deepseek-R1Guardrail=Llama Guard2026.0167.367.368.267.369.267.56.44.35.1
Deepseek-R1Guardrail=OpenAI2026.0166.560.56156.262.2585.84.34.5
Deepseek-R1Guardrail=Perspective2026.0166.266.864.765.865.263.46.455.1
Grok4Guardrail=Llama Guard2026.0165.867.765.567.370.366.26.677.6
Gemini 2.5-ProGuardrail=Llama Guard2026.0164.759.854.861.361.564.25.33.36.4
Llama4Guardrail=Llama Guard2026.0164.366.866.56565.267.56.678.2
Grok4Guardrail=OpenAI2026.0163.863.55954.859.5606.466.6
Llama4Guardrail=OpenAI2026.016362.860.259.862.864.27.67.26.6
Gemini 2.5-ProGuardrail=Perspective2026.0161.761.360.259.561.363.24.53.96
Gemini 2.5-ProGuardrail=OpenAI2026.0157.856.85655.35955.94.53.56
GPT4Guardrail=Perspective2026.0157.265.566.263.864.762.76.63.34.5
ClaudeGuardrail=None2026.015149524759.857.46.64.35.6
ClaudeGuardrail=OpenAI2026.0150.546.350.54358495.33.13.7
ClaudeGuardrail=Perspective2026.0148.547.247.245.858.352.95.33.74.3
ClaudeGuardrail=Llama Guard2026.0142.845.544.843.352465.83.34.5