Safety Detection on MAS LLaMA-3.1-8B (Current)
99.15AccuracyAgentLens
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AgentLensMethod Category=Interpretability-based2026.06 | 99.15 | 99.17 | |
| Gemini-3-ProMethod Category=LLM-as-a-Judge2026.06 | 96.16 | 96.28 | |
| Claude Sonnet-4.6Method Category=LLM-as-a-Judge2026.06 | 95.31 | 95.58 | |
| LLaMA-Guard-4-12BMethod Category=Guardrail-based2026.06 | 94.03 | 93.86 | |
| Deepseek-V4-FlashMethod Category=LLM-as-a-Judge2026.06 | 93.18 | 93.31 | |
| Claude Opus-4.6Method Category=LLM-as-a-Judge2026.06 | 92.96 | 93.52 | |
| Deepseek-V4-ProMethod Category=LLM-as-a-Judge2026.06 | 91.04 | 91.18 | |
| AGrail (GPT-4o)Method Category=Guardrail-based2026.06 | 82.09 | 84.55 | |
| GPT-4oMethod Category=LLM-as-a-Judge2026.06 | 78.04 | 82.02 | |
| GPT-5.2Method Category=LLM-as-a-Judge2026.06 | 68.02 | 76.27 |