Image Difference Captioning on Spot-the-Diff (test)
14.7METEORFINER
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| FINERMethod category=LLM-based2026.03 | 14.7 | 12.9 | 61.8 | 35.5 | |
| MCT-CCDiffMethod category=Non-LLM-based2026.03 | 14.5 | 10.8 | 41.7 | 35.5 | |
| RDD+ACRMethod category=Non-LLM-based2026.03 | 13.9 | 9.2 | 43.6 | 31 | |
| DIRL+CCRMethod category=Non-LLM-based2026.03 | 13.8 | 10.3 | 40.9 | 32.8 | |
| ProCapMethod category=Non-LLM-based2026.03 | 13.6 | 11 | 42.7 | 33.7 | |
| SMARTMethod category=Non-LLM-based2026.03 | 13.5 | — | 39.4 | 31.6 | |
| MGM-7B + RPBackbone=MGM-7B, Object Replacement (RP) Fine-tuning=true2024.08 | 13.1 | 10.8 | 53.5 | 33 | |
| MURAT+GCMMethod category=Non-LLM-based2026.03 | 13.1 | 10.2 | 39.4 | 33.1 | |
| LLaVA-1.5-7B + RPBackbone=LLaVA-1.5-7B, Object Replacement (RP) Fine-tuning=true2024.08 | 13 | 9.7 | 43.2 | 30.8 | |
| LLaVA-1.5+RPMethod category=LLM-based2026.03 | 13 | 9.7 | 43.2 | 30.8 | |
| InternVL2-8B + RPBackbone=InternVL2-8B, Object Replacement (RP) Fine-tuning=true2024.08 | 12.8 | 8.4 | 32.2 | 28.5 | |
| VACC2024.08 | 12.6 | 9.7 | 41.5 | 32.1 | |
| DUDA+Aux2024.08 | 12.5 | 8.1 | 34.5 | 29.9 | |
| VAM2024.08 | 12.4 | 10.1 | 38.1 | 31.3 | |
| DUDA+AuxMethod category=Non-LLM-based2026.03 | 12.4 | 8.1 | 38.1 | 31.3 | |
| SCORER+CBRMethod category=Non-LLM-based2026.03 | 12.2 | 10.2 | 38.9 | — | |
| LLaVA-1.5-7BBackbone=LLaVA-1.5-7B2024.08 | 12 | 8.5 | 38.3 | 30.1 | |
| MGM-7BBackbone=MGM-7B2024.08 | 12 | 9.9 | 46.3 | 31.5 | |
| DUDAMethod category=Non-LLM-based2026.03 | 11.8 | 9.1 | 32.5 | 29.1 | |
| IFDC2024.08 | 11.7 | 8.7 | 37 | 30.2 | |
| InternVL2-8B-FTBackbone=InternVL2-8B, Fine-tuning Strategy=FT (Official baseline)2024.08 | 11.7 | 6.6 | 26.5 | 27.3 | |
| IFDCMethod category=Non-LLM-based2026.03 | 11.7 | 8.7 | 37 | 30.2 |