Fine-grained risk diagnosis on ATBench
75.2Risk Source ScoreAgentDoG 1.5-4B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| AgentDoG 1.5-4BModel Category=Our Models, Size=4B2026.05 | 75.2 | 27.5 | 62.9 | 55.2 | |
| AgentDoG 1.5-8BModel Category=Our Models, Size=8B2026.05 | 72.9 | 24.6 | 52.5 | 50 | |
| AgentDoG 1.5-2BModel Category=Our Models, Size=2B2026.05 | 68 | 24 | 53.8 | 48.6 | |
| AgentDoG 1.5-0.8BModel Category=Our Models, Size=0.8B2026.05 | 65.7 | 18.4 | 44.9 | 43 | |
| AgentDoG 1.0-4BModel Category=Our Models2026.05 | 46.8 | 16.5 | 40.6 | 34.6 | |
| GPT-5.4Model Category=Closed-Source Models2026.05 | 33.6 | 13.5 | 30.2 | 25.8 | |
| GPT-5.2Model Category=Closed-Source Models2026.05 | 29.5 | 12 | 26.8 | 22.8 | |
| Gemini-3.1-ProModel Category=Closed-Source Models2026.05 | 24.8 | 12.6 | 18.5 | 18.6 | |
| AgentDoG 1.5-4B-UModel Category=Our Models, Size=4B, Type=Unified2026.05 | 24.1 | 9.5 | 28.4 | 20.7 | |
| Gemini-3-FlashModel Category=Closed-Source Models2026.05 | 18.4 | 8.3 | 15 | 13.9 | |
| QwQ-32BModel Category=Open-Source Models2026.05 | 15.8 | 9.4 | 22.9 | 16 | |
| Qwen3.5-397BModel Category=Open-Source Models2026.05 | 7.7 | 3.6 | 6.8 | 6 | |
| Qwen3.5-2BModel Category=Open-Source Models2026.05 | 7.7 | 6.6 | 11.1 | 8.5 | |
| Qwen3-235BModel Category=Open-Source Models2026.05 | 7 | 11.6 | 26.6 | 15.1 | |
| Qwen3.5-4BModel Category=Open-Source Models2026.05 | 6.6 | 3 | 8.2 | 5.9 | |
| Llama3.1-8B-Inst.Model Category=Open-Source Models2026.05 | 6.2 | 5.8 | 15.5 | 9.2 | |
| Qwen2.5-7B-Inst.Model Category=Open-Source Models2026.05 | 5.3 | 6 | 15.5 | 8.9 | |
| Qwen3.5-0.8BModel Category=Open-Source Models2026.05 | 1.3 | 2.9 | 4.7 | 3 | |
| Qwen3-4B-Inst.Model Category=Open-Source Models2026.05 | 1 | 9.6 | 21.2 | 10.6 |