Artifact Localization on ArtiBench (test)
0.119mIoUInternVL3.5-8B + ArtiAgent
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| InternVL3.5-8B + ArtiAgentMode=Fine-tuned with ArtiAgent, Training Set Size=100K2026.02 | 0.119 | 0.176 | |
| Qwen2.5-VL-7B + ArtiAgentMode=Fine-tuned with ArtiAgent, Training Set Size=100K2026.02 | 0.111 | 0.168 | |
| Gemini-2.5-ProSystem Category=Proprietary VLM2026.02 | 0.095 | 0.147 | |
| DiffDoctorMethod Type=Artifact Segmentation Algorithm2026.02 | 0.081 | 0.137 | |
| LEGIONMethod Type=Artifact Segmentation Algorithm2026.02 | 0.062 | 0.099 | |
| GPT-5System Category=Proprietary VLM2026.02 | 0.061 | 0.099 | |
| GPT-4oSystem Category=Proprietary VLM2026.02 | 0.049 | 0.084 | |
| PALMethod Type=Artifact Segmentation Algorithm2026.02 | 0.04 | 0.066 | |
| Qwen2.5-VL-7BMode=Vanilla2026.02 | 0.01 | 0.014 | |
| InternVL3.5-8BMode=Vanilla2026.02 | 0.01 | 0.015 |