List your company

Jailbreak Detection on GoalFrameBench

94AccuracyFrameShield-Crit
56.5666.287685.72Feb 23, 2026

Evaluation results

MethodLinks
FrameShield-CritLLM Backbone=Qwen2.5-7B2026.029493
FrameShield-CritLLM Backbone=Qwen3-4B2026.028987
FrameShield-LastLLM Backbone=Llama3-8B2026.028279
FrameShield-CritLLM Backbone=Vicuna-7B2026.028179
FrameShield-CritLLM Backbone=Vicuna-13B2026.028182
FrameShield-CritLLM Backbone=Llama2-7B2026.028079
FrameShield-LastLLM Backbone=Mistral-7B2026.027974
FrameShield-LastLLM Backbone=Qwen2.5-7B2026.027976
FrameShield-LastLLM Backbone=Vicuna-13B2026.027679
FrameShield-LastLLM Backbone=Qwen3-4B2026.027675
FrameShield-CritLLM Backbone=Llama3-8B2026.027669
FrameShield-CritLLM Backbone=Mistral-7B2026.027669
JBShieldLLM Backbone=Llama3-8B2026.027575
FrameShield-CritLLM Backbone=Qwen2-0.5B2026.027465
JBShieldLLM Backbone=Llama2-7B2026.027376
JBShieldLLM Backbone=Mistral-7B2026.027175
JBShieldLLM Backbone=Qwen3-4B2026.027074
FrameShield-LastLLM Backbone=Vicuna-7B2026.026775
JBShieldLLM Backbone=Qwen2.5-7B2026.026668
FrameShield-LastLLM Backbone=Qwen2-0.5B2026.026674
JBShieldLLM Backbone=Vicuna-13B2026.026367
JBShieldLLM Backbone=Vicuna-7B2026.026063
JBShieldLLM Backbone=Qwen2-0.5B2026.025962
FrameShield-LastLLM Backbone=Llama2-7B2026.025870