Safety Detection on MAS Qwen-2.5-7B (Lookahead)
96.77AccuracyAgentLens
Evaluation Results
| Method | Links | |
|---|---|---|
| AgentLensMethod Category=Interpretability-based2026.06 | 96.77 | |
| GPT-4oMethod Category=LLM-as-a-Judge2026.06 | 87.1 | |
| Claude Opus-4.6Method Category=LLM-as-a-Judge2026.06 | 87.1 | |
| Claude Sonnet-4.6Method Category=LLM-as-a-Judge2026.06 | 83.87 | |
| GPT-5.2Method Category=LLM-as-a-Judge2026.06 | 64.52 | |
| Gemini-3-ProMethod Category=LLM-as-a-Judge2026.06 | 35.48 | |
| Deepseek-V4-FlashMethod Category=LLM-as-a-Judge2026.06 | 35.48 | |
| Deepseek-V4-ProMethod Category=LLM-as-a-Judge2026.06 | 29.03 |