Multimodal Understanding on MM-Vet OOD
96.33AccuracyVLMShield
Evaluation results
| Method | Links | |
|---|---|---|
| VLMShieldBase VLM=LLaVA-1.5-13B2026.04 | 96.33 | |
| VLMShieldBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 96.33 | |
| VLMGuardBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 96.08 | |
| ASTRABase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 95.8 | |
| VLMGuardBase VLM=LLaVA-1.5-13B2026.04 | 95 | |
| JailGuardBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 94.38 | |
| ASTRABase VLM=LLaVA-1.5-13B2026.04 | 93.54 | |
| CIDERBase VLM=LLaVA-1.5-13B2026.04 | 93.28 | |
| CIDERBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 93.28 | |
| ECSOBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 93.23 | |
| JailGuardBase VLM=LLaVA-1.5-13B2026.04 | 89.45 | |
| MirrorCheckBase VLM=LLaVA-1.5-13B2026.04 | 89.41 | |
| MirrorCheckBase VLM=Qwen2.5-VL-7B-Instruct2026.04 | 89.41 | |
| ECSOBase VLM=LLaVA-1.5-13B2026.04 | 89.04 |