Industrial Anomaly Detection on MPDD (Balanced Accuracy)
72.7Balanced AccuracyIndusAgent
Evaluation Results
| Method | Links | |
|---|---|---|
| IndusAgentType=Open Source, Param.=8B, Evaluation Protocol=Zero-shot2026.05 | 72.7 | |
| Anomaly-OVType=Open Source, Param.=7B, Evaluation Protocol=Zero-shot2026.05 | 70.3 | |
| GPT-4.1-miniType=Commercial, Param.=/, Evaluation Protocol=Zero-shot2026.05 | 69.8 | |
| GPT-4o-miniType=Commercial, Param.=/, Evaluation Protocol=Zero-shot2026.05 | 67.9 | |
| GPT-4.1Type=Commercial, Param.=/, Evaluation Protocol=Zero-shot2026.05 | 66.7 | |
| Claude-Sonnet-4Type=Commercial, Param.=/, Evaluation Protocol=Zero-shot2026.05 | 65.9 | |
| Qwen2.5-VL-InstructType=Open Source, Param.=72B, Evaluation Protocol=Zero-shot2026.05 | 64.7 | |
| GPT-4.1-nanoType=Commercial, Param.=/, Evaluation Protocol=Zero-shot2026.05 | 61.7 | |
| LLaVA-1.5Type=Open Source, Param.=13B, Evaluation Protocol=Zero-shot2026.05 | 61.4 | |
| GPT-4oType=Commercial, Param.=/, Evaluation Protocol=Zero-shot2026.05 | 60.3 | |
| InternVL-2.5Type=Open Source, Param.=4B, Evaluation Protocol=Zero-shot2026.05 | 59.1 | |
| AnomalyR1Type=Open Source, Param.=3B, Evaluation Protocol=Zero-shot2026.05 | 56 | |
| Qwen2.5-VL-InstructType=Open Source, Param.=7B, Evaluation Protocol=Zero-shot2026.05 | 56 | |
| AnomalyGPTType=Open Source, Param.=7B, Evaluation Protocol=Zero-shot2026.05 | 54.2 | |
| Qwen2.5-VL-InstructType=Open Source, Param.=3B, Evaluation Protocol=Zero-shot2026.05 | 52.9 | |
| LLaVA-OneVision-SIType=Open Source, Param.=0.5B, Evaluation Protocol=Zero-shot2026.05 | 50 | |
| Anomaly-OVType=Open Source, Param.=0.5B, Evaluation Protocol=Zero-shot2026.05 | 50 | |
| Qwen3-VL-InstructType=Open Source, Param.=8B, Evaluation Protocol=Zero-shot2026.05 | 50 | |
| LLaVA-1.6Type=Open Source, Param.=34B, Evaluation Protocol=Zero-shot2026.05 | 50 |