List your company

Safety Evaluation on Malicious Instruct

0ASRChat model
-3.92422.56349.0575.537Jan 7, 2026Jan 25, 2026Feb 13, 2026Mar 4, 2026Mar 23, 2026Apr 11, 2026Apr 30, 2026

Evaluation results

MethodLinks
Chat modelBackbone LLM=LLaMa-2 7B2026.01040
ITIBackbone LLM=Qwen-2.5 7B2026.01090
Chat modelBackbone LLM=Qwen-2.5 7B2026.01733
ARREST-SCBackbone LLM=LLaMa-3.1 8B2026.011085
ARREST-SCBackbone LLM=Qwen-2.5 7B2026.011585
ARREST-SCBackbone LLM=LLaMa-2 7B2026.011999
Chat modelBackbone LLM=LLaMa-3.1 8B2026.011963
ARREST-SBBackbone LLM=Qwen-2.5 7B2026.011979
Chat modelBackbone LLM=Yi-1.5 9B2026.012074
ITIBackbone LLM=LLaMa-3.1 8B2026.012270
ARREST-SBBackbone LLM=LLaMa-3.1 8B2026.012478
Base ModelBackbone LLM=LLaMa-3.1 8B2026.012652
Base ModelBackbone LLM=Qwen-2.5 7B2026.012930
ARREST-SBBackbone LLM=LLaMa-2 7B2026.013095
ARREST-SCBackbone LLM=Yi-1.5 9B2026.013791
ARREST-SBBackbone LLM=Yi-1.5 9B2026.014790.1
CEBackbone=PHI-4-14B2026.0449.3—
TOFUBackbone=PHI-4-14B2026.0452.4—
FLBackbone=PHI-4-14B2026.0452.8—
Base ModelBackbone LLM=LLaMa-2 7B2026.015337
ITIBackbone LLM=LLaMa-2 7B2026.015372
ITIBackbone LLM=Yi-1.5 9B2026.015365
Base ModelBackbone LLM=Yi-1.5 9B2026.015441
FLBackbone=QWEN-3-8B2026.0454.3—
CEBackbone=QWEN-3-8B2026.0454.4—
GEMBackbone=PHI-4-14B2026.0456.7—
TOFUBackbone=QWEN-3-8B2026.0456.8—
GEMBackbone=QWEN-3-8B2026.0461.9—
CEBackbone=MISTRAL-12B2026.0486—
TOFUBackbone=MISTRAL-12B2026.0487.2—
FLBackbone=MISTRAL-12B2026.0488.4—
GEMBackbone=MISTRAL-12B2026.0490—
FLBackbone=LLAMA-3.1-8B2026.0494—
CEBackbone=LLAMA-3.1-8B2026.0494.7—
TOFUBackbone=OLMO-2-13B2026.0494.9—
GEMBackbone=LLAMA-3.1-8B2026.0495.2—
FLBackbone=OLMO-2-13B2026.0495.3—
TOFUBackbone=LLAMA-3.1-8B2026.0495.3—
GEMBackbone=OLMO-2-13B2026.0496.6—
CEBackbone=OLMO-2-13B2026.0496.8—
TOFUBackbone=PYTHIA-12B2026.0497—
FLBackbone=PYTHIA-12B2026.0497.1—
GEMBackbone=PYTHIA-12B2026.0497.2—
CEBackbone=PYTHIA-12B2026.0498.1—