Risk Identification on EARBench
94.73RIRHomeGuard-8B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HomeGuard-8BModel Category=Ours, Parameters=8B2026.03 | 94.73 | 72.12 | |
| GPT-4o-miniModel Category=Proprietary Models2026.03 | 91.21 | 65.35 | |
| HomeGuard-4BModel Category=Ours, Parameters=4B2026.03 | 90.06 | 64.78 | |
| Qwen3-VL-235B-A22B-ThinkingModel Category=Open-Sourced VLMs, Parameters=235B-A22B2026.03 | 89.83 | 67.8 | |
| Gemini-3-ProModel Category=Proprietary Models2026.03 | 89.08 | 65.35 | |
| RoboBrain2.5-8BModel Category=Open-Sourced VLMs, Parameters=8B2026.03 | 88.7 | 60.26 | |
| AgentSpecModel Category=Runtime Safeguard2026.03 | 88.51 | 61.77 | |
| Gemini-2.5-flashModel Category=Proprietary Models2026.03 | 88.32 | 67.23 | |
| Qwen3-VL-8B-ThinkingModel Category=Open-Sourced VLMs, Parameters=8B2026.03 | 86.06 | 57.25 | |
| Qwen3-VL-4B-ThinkingModel Category=Open-Sourced VLMs, Parameters=4B2026.03 | 82.67 | 52.35 | |
| ThinkSafeModel Category=Runtime Safeguard2026.03 | 66.1 | 51.79 | |
| PoexModel Category=Runtime Safeguard2026.03 | 64.78 | 42.75 |