List your company

Safety Evaluation on Jailbreak Bench

3ASRChat model
-0.221.44364.6Jan 7, 2026Jan 17, 2026Jan 27, 2026Feb 7, 2026Feb 17, 2026Feb 27, 2026Mar 10, 2026

Evaluation results

MethodLinks
Chat modelBackbone LLM=LLaMa-2 7B2026.01337
SafeReActModel=Finance-Llama3.1-8B-Instruct2026.033—
Chat modelBackbone LLM=Qwen-2.5 7B2026.011062
ITIBackbone LLM=Qwen-2.5 7B2026.011056
Chat modelBackbone LLM=LLaMa-3.1 8B2026.011377.7
ARREST-SCBackbone LLM=LLaMa-3.1 8B2026.011785
ARREST-SCBackbone LLM=Qwen-2.5 7B2026.012280
ITIBackbone LLM=LLaMa-3.1 8B2026.012670
ARREST-SBBackbone LLM=LLaMa-3.1 8B2026.012979
Chat modelBackbone LLM=Yi-1.5 9B2026.013070
ARREST-SBBackbone LLM=Qwen-2.5 7B2026.013679
ARREST-SCBackbone LLM=LLaMa-2 7B2026.014294
Base ModelBackbone LLM=LLaMa-3.1 8B2026.014270
Base ModelBackbone LLM=Qwen-2.5 7B2026.014255
OriginalModel=Finance-Llama3.1-8B-Instruct2026.0347—
ARREST-SCBackbone LLM=Yi-1.5 9B2026.015275
ARREST-SBBackbone LLM=LLaMa-2 7B2026.015489
ARREST-SBBackbone LLM=Yi-1.5 9B2026.016380
ITIBackbone LLM=Yi-1.5 9B2026.016470
ITIBackbone LLM=LLaMa-2 7B2026.017577
Base ModelBackbone LLM=Yi-1.5 9B2026.017735
Base ModelBackbone LLM=LLaMa-2 7B2026.018331