Safety Evaluation on Malicious Instruct
0ASRChat model
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Chat modelBackbone LLM=LLaMa-2 7B2026.01 | 0 | 40 | |
| ITIBackbone LLM=Qwen-2.5 7B2026.01 | 0 | 90 | |
| Chat modelBackbone LLM=Qwen-2.5 7B2026.01 | 7 | 33 | |
| ARREST-SCBackbone LLM=LLaMa-3.1 8B2026.01 | 10 | 85 | |
| ARREST-SCBackbone LLM=Qwen-2.5 7B2026.01 | 15 | 85 | |
| ARREST-SCBackbone LLM=LLaMa-2 7B2026.01 | 19 | 99 | |
| Chat modelBackbone LLM=LLaMa-3.1 8B2026.01 | 19 | 63 | |
| ARREST-SBBackbone LLM=Qwen-2.5 7B2026.01 | 19 | 79 | |
| Chat modelBackbone LLM=Yi-1.5 9B2026.01 | 20 | 74 | |
| ITIBackbone LLM=LLaMa-3.1 8B2026.01 | 22 | 70 | |
| ARREST-SBBackbone LLM=LLaMa-3.1 8B2026.01 | 24 | 78 | |
| Base ModelBackbone LLM=LLaMa-3.1 8B2026.01 | 26 | 52 | |
| Base ModelBackbone LLM=Qwen-2.5 7B2026.01 | 29 | 30 | |
| ARREST-SBBackbone LLM=LLaMa-2 7B2026.01 | 30 | 95 | |
| ARREST-SCBackbone LLM=Yi-1.5 9B2026.01 | 37 | 91 | |
| ARREST-SBBackbone LLM=Yi-1.5 9B2026.01 | 47 | 90.1 | |
| CEBackbone=PHI-4-14B2026.04 | 49.3 | — | |
| TOFUBackbone=PHI-4-14B2026.04 | 52.4 | — | |
| FLBackbone=PHI-4-14B2026.04 | 52.8 | — | |
| Base ModelBackbone LLM=LLaMa-2 7B2026.01 | 53 | 37 | |
| ITIBackbone LLM=LLaMa-2 7B2026.01 | 53 | 72 | |
| ITIBackbone LLM=Yi-1.5 9B2026.01 | 53 | 65 | |
| Base ModelBackbone LLM=Yi-1.5 9B2026.01 | 54 | 41 | |
| FLBackbone=QWEN-3-8B2026.04 | 54.3 | — | |
| CEBackbone=QWEN-3-8B2026.04 | 54.4 | — | |
| GEMBackbone=PHI-4-14B2026.04 | 56.7 | — | |
| TOFUBackbone=QWEN-3-8B2026.04 | 56.8 | — | |
| GEMBackbone=QWEN-3-8B2026.04 | 61.9 | — | |
| CEBackbone=MISTRAL-12B2026.04 | 86 | — | |
| TOFUBackbone=MISTRAL-12B2026.04 | 87.2 | — | |
| FLBackbone=MISTRAL-12B2026.04 | 88.4 | — | |
| GEMBackbone=MISTRAL-12B2026.04 | 90 | — | |
| FLBackbone=LLAMA-3.1-8B2026.04 | 94 | — | |
| CEBackbone=LLAMA-3.1-8B2026.04 | 94.7 | — | |
| TOFUBackbone=OLMO-2-13B2026.04 | 94.9 | — | |
| GEMBackbone=LLAMA-3.1-8B2026.04 | 95.2 | — | |
| FLBackbone=OLMO-2-13B2026.04 | 95.3 | — | |
| TOFUBackbone=LLAMA-3.1-8B2026.04 | 95.3 | — | |
| GEMBackbone=OLMO-2-13B2026.04 | 96.6 | — | |
| CEBackbone=OLMO-2-13B2026.04 | 96.8 | — | |
| TOFUBackbone=PYTHIA-12B2026.04 | 97 | — | |
| FLBackbone=PYTHIA-12B2026.04 | 97.1 | — | |
| GEMBackbone=PYTHIA-12B2026.04 | 97.2 | — | |
| CEBackbone=PYTHIA-12B2026.04 | 98.1 | — |