List your company

Jailbreak Detection on SafeDialBench

0.912AUCSAFEDREAM
0.679040.739520.80.86048Apr 18, 2026

Evaluation results

MethodLinks
SAFEDREAMMethod category=Multi-turn stateful detectors (external), Target LLM=Qwen2.5-7B-Instruct2026.040.91290.16.61.2
TARSMethod category=Alignment-based methods (modify LLM weights), Target LLM=Qwen2.5-7B-Instruct2026.040.86584.75.70.55
X-BoundaryMethod category=Alignment-based methods (modify LLM weights), Target LLM=Qwen2.5-7B-Instruct2026.040.85883.99.40.37
Circuit BreakerMethod category=Alignment-based methods (modify LLM weights), Target LLM=Qwen2.5-7B-Instruct2026.040.83182.37.10.42
SelfDefendMethod category=external, stateless, Target LLM=Qwen2.5-7B-Instruct2026.040.73472.111.50.18
BingoGuardMethod category=external, stateless, Target LLM=Qwen2.5-7B-Instruct2026.040.7216813.20.55
GuardReasonerMethod category=external, stateless, Target LLM=Qwen2.5-7B-Instruct2026.040.7197313.70.23
R2-GuardMethod category=external, stateless, Target LLM=Qwen2.5-7B-Instruct2026.040.69169.816.90.64
WildGuardMethod category=external, stateless, Target LLM=Qwen2.5-7B-Instruct2026.040.68862.517.70.89