Multimodal Understanding on GPT4V-Caption (IOD)
100AccuracyVLMShield
Evaluation results
| Method | Links | |
|---|---|---|
| VLMShieldBase VLM=LLaVA-1.5-13B2026.04 | 100 | |
| VLMShieldBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 100 | |
| CIDERBase VLM=LLaVA-1.5-13B2026.04 | 97.8 | |
| CIDERBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 97.8 | |
| ASTRABase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 97.74 | |
| JailGuardBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 97.36 | |
| VLMGuardBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 97.33 | |
| ECSOBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 96.3 | |
| ASTRABase VLM=LLaVA-1.5-13B2026.04 | 96.15 | |
| VLMGuardBase VLM=LLaVA-1.5-13B2026.04 | 95.24 | |
| JailGuardBase VLM=LLaVA-1.5-13B2026.04 | 95.09 | |
| ECSOBase VLM=LLaVA-1.5-13B2026.04 | 93.98 | |
| MirrorCheckBase VLM=LLaVA-1.5-13B2026.04 | 92.06 | |
| MirrorCheckBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 92.06 |