Visual Anomaly Detection on CrossDomainVAD-12 (test)
97.5D1 ScoreAnomalyDINO
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AnomalyDINOCategory=Specialist non-VLM AD methods, Protocol=4-shot2026.05 | 97.5 | 88.9 | 64.2 | 53 | 84.2 | 72.7 | 68.7 | 69.5 | 73.9 | 56.6 | 63.5 | 98.2 | 74.2 | |
| SubspaceADCategory=Specialist non-VLM AD methods, Protocol=4-shot2026.05 | 96.6 | 91.9 | 70.1 | 56.5 | 84.1 | 75.7 | 56.6 | 68.8 | 81.6 | 71 | 60.8 | 98.9 | 76 | |
| AD-CopilotCategory=Specialist VLM, Protocol=4-shot2026.05 | 96.1 | 87.3 | 57.4 | 53.9 | 69.2 | 68.8 | 58.9 | 65.7 | 71.2 | 57.3 | 74.8 | 79.3 | 70 | |
| AnomalyClawCategory=Our agent, Backbone=Qwen3.5-VL-27B, Protocol=4-shot2026.05 | 95.6 | 77.3 | 67.3 | 51.7 | 64.9 | 67.2 | 84.2 | 63 | 84.1 | 61.6 | 81.7 | 99 | 74.8 | |
| AnomalyClawCategory=Our agent, Backbone=GPT-5.5, Protocol=4-shot2026.05 | 95.5 | 91.2 | 77 | 55 | 84.4 | 77.8 | 86.7 | 74.6 | 83.9 | 68.3 | 84.9 | 96.7 | 81.4 | |
| GPT-5.5Category=General VLM, Backbone=GPT-5.5, Protocol=4-shot2026.05 | 94.3 | 89.3 | 66.7 | 53.6 | 82.6 | 57.3 | 87.4 | 68.5 | 78.7 | 48.9 | 77.7 | 96.1 | 75.1 | |
| Qwen3.5-VL-27B (JSON)Category=General VLM, Backbone=Qwen3.5-VL-27B, Protocol=4-shot, Format=JSON2026.05 | 90.6 | 76.6 | 60.7 | 48.2 | 68.4 | 58 | 84.2 | 55.4 | 78.7 | 52.2 | 86 | 96.4 | 71.3 | |
| AnomalyClawCategory=Our agent, Backbone=Seed2.0-lite, Protocol=4-shot2026.05 | 89.1 | 93.2 | 81.9 | 53.6 | 87.8 | 70.6 | 79.9 | 67.3 | 68.5 | 56 | 75.9 | 96.9 | 76.7 | |
| AnomalyVFMCategory=Specialist non-VLM AD methods, Protocol=0-shot2026.05 | 89 | 73.6 | 57.2 | 47.8 | 56.6 | 79.8 | 30.6 | 57.2 | 79.1 | 57.7 | 54 | 44.3 | 60.6 | |
| VisualADCategory=Specialist non-VLM AD methods, Protocol=0-shot2026.05 | 88.2 | 93.6 | 59.4 | 49.7 | 70.4 | 60.3 | 34.5 | 73 | 68.9 | 67.5 | 54.7 | 72.5 | 66.1 | |
| Seed2.0-liteCategory=General VLM, Backbone=Seed2.0-lite, Protocol=4-shot2026.05 | 84.1 | 88.9 | 68.3 | 53 | 87.2 | 44.5 | 75.4 | 63 | 61.6 | 46.5 | 73.1 | 79.8 | 68.8 | |
| IAD-R1Category=Specialist VLM, Protocol=0-shot2026.05 | 81.1 | 78.2 | 47.7 | 49.6 | 58.7 | 65 | 23 | 50.5 | 47 | 75.2 | 44.9 | 21.4 | 53.5 |