List your company

Toxic Comment Classification on Toxic Comment

401.1Average TokensTopic
-3.46101.57206.6311.63Jan 15, 2026

Evaluation results

MethodLinks
TopicVictim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.01401.110099.999.999.7
Random Criteria (Ablation)Victim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.0120170.768.272.273
Criteria Attack (Double Criteria)Victim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.01200.393.794.295.794.2
Single Criteria (Ablation)Victim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.01165.591.992.294.892.3
No Fake Reasoning (Ablation)Victim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.0154.754.855.349.757.2
Separator InjectionVictim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.0148.978.755.357.486.6
CombinedVictim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.012910025.665.8100
Fake CompletionVictim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.012338.827.733.166.1
IgnoreVictim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.0118.198.74.71597.7
Escape SeparationVictim Model=Gemma-3-27B, Attacker Model=Qwen3-30B2026.0112.147.727.829.545.5