Anomaly Reasoning on MVTec-COT (test)
27.1ROUGE-LQwen2.5-VL+CGRO*
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-VL+CGRO*Param.=7B, Anno.=I, ReAL=true2026.03 | 27.1 | 74.7 | |
| Qwen2.5-VL+R1*Param.=7B, Anno.=I, ReAL=true2026.03 | 26.3 | 73.8 | |
| Qwen2.5-VL+CGRO*Param.=3B, Anno.=I, ReAL=true2026.03 | 25.6 | 72.4 | |
| Qwen2.5-VL+R1*Param.=3B, Anno.=I, ReAL=true2026.03 | 24.2 | 71.7 | |
| LLaVA-OneVision-OVParam.=7B, Anno.=/2026.03 | 23.7 | 70.5 | |
| InternVL-3Param.=78B, Anno.=/2026.03 | 22.2 | 70.1 | |
| MiniCPM-V-2.6Param.=8B, Anno.=/2026.03 | 21.7 | 68.3 | |
| Qwen2.5-VLParam.=72B, Anno.=/2026.03 | 21.1 | 67.5 | |
| EIADParam.=7B, Anno.=T+I2026.03 | 21.1 | 65.7 | |
| GPT-4.1Param.=/, Anno.=/2026.03 | 20.8 | 69.9 | |
| Qwen2.5-VLParam.=7B, Anno.=/2026.03 | 20.8 | 65.5 | |
| Qwen2.5-VLParam.=32B, Anno.=/2026.03 | 20 | 66.1 | |
| GPT-4oParam.=/, Anno.=/2026.03 | 19.9 | 60.4 | |
| InternVL-2.5Param.=8B, Anno.=/2026.03 | 19.7 | 57.5 | |
| Qwen3VL-thinkingParam.=8B, Anno.=/2026.03 | 19.3 | 61.9 | |
| GPT-4o-miniParam.=/, Anno.=/2026.03 | 19 | 65.2 | |
| LLaVA-1.6Param.=7B, Anno.=/2026.03 | 18.6 | 54.2 | |
| Gemini-2.5-FlashParam.=/, Anno.=/2026.03 | 17.6 | 68.1 | |
| Qwen3VL-thinking-A3BParam.=30B, Anno.=/2026.03 | 16.9 | 58.9 | |
| Qwen2.5-VLParam.=3B, Anno.=/2026.03 | 16.8 | 49 | |
| GPT-5-nanoParam.=/, Anno.=/2026.03 | 14 | 50 | |
| AnomalyGPTParam.=7B, Anno.=T+I+P2026.03 | 11.9 | 36.7 | |
| LLaVA-OneVision-SIParam.=7B, Anno.=/2026.03 | 10.5 | 39.1 | |
| TriadParam.=7B, Anno.=T+I2026.03 | 8.6 | 35.9 |