Robustness to Natural Adversarial Examples on NaturalBench
9.89AccuracyDINOv2
Evaluation results
| Method | Links | |
|---|---|---|
| DINOv2LLM=Qwen2-7B2026.02 | 9.89 | |
| DINOv2LLM=Phi22026.02 | 8.42 | |
| mFRESALLM=Qwen2-7B2026.02 | 6.84 | |
| DINOSAURv2LLM=Qwen2-7B2026.02 | 6.16 | |
| VQDINOMLPLLM=Qwen2-7B2026.02 | 6.07 | |
| FT-DINOSAURLLM=Qwen2-7B2026.02 | 5.42 | |
| StableLSDLLM=Qwen2-7B2026.02 | 5.21 | |
| mFRESALLM=Phi22026.02 | 4.21 | |
| DINOSAURLLM=Qwen2-7B2026.02 | 3.95 | |
| VQDINOMLPLLM=Phi22026.02 | 3.89 | |
| Slot DiffusionLLM=Qwen2-7B2026.02 | 3.74 | |
| SPOTLLM=Qwen2-7B2026.02 | 3.68 | |
| DINOSAURv2LLM=Phi22026.02 | 3.37 | |
| StableLSDLLM=Phi22026.02 | 3 | |
| FT-DINOSAURLLM=Phi22026.02 | 2.89 | |
| SPOTLLM=Phi22026.02 | 2.42 | |
| Slot DiffusionLLM=Phi22026.02 | 2.21 | |
| DINOSAURLLM=Phi22026.02 | 1.89 | |
| Blind-VLMLLM=Qwen2-7B2026.02 | 0.52 | |
| Blind-VLMLLM=Phi22026.02 | 0.42 |