Image Difference Captioning on Image-Edit-Request (test)
16.6BLEUMGM-7B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MGM-7Btraining_data=Object Replacement (RP)2024.08 | 16.6 | 18.2 | 68.1 | 45.7 | |
| MGM-7B2024.08 | 16.5 | 17.7 | 66.8 | 44.8 | |
| LLaVA-1.5-7Btraining_data=Object Replacement (RP)2024.08 | 16.2 | 19.5 | 60.9 | 46.7 | |
| LLaVA-1.5+RPMethod category=LLM-based2026.03 | 16.2 | 19.5 | 60.9 | 46.7 | |
| LLaVA-1.5-7B2024.08 | 15.1 | 17.8 | 60.6 | 45.2 | |
| FINERMethod category=LLM-based2026.03 | 13.3 | 14.6 | 50.5 | 39.6 | |
| InternVL2-8Btraining_data=Object Replacement (RP)2024.08 | 12.5 | 14.2 | 56 | 39.4 | |
| InternVL2-8B-FT2024.08 | 12.4 | 14.1 | 51.5 | 38.9 | |
| ProCapMethod category=Non-LLM-based2026.03 | 11.7 | 15.9 | 40.6 | 43.2 | |
| DIRL+CCRMethod category=Non-LLM-based2026.03 | 10.9 | 15 | 34.1 | 41 | |
| SMARTMethod category=Non-LLM-based2026.03 | 10.5 | 15.2 | 37.8 | 39.1 | |
| MCT-CCDiffMethod category=Non-LLM-based2026.03 | 10.2 | 15.4 | 38.3 | 41.2 | |
| VARD2024.08 | 10 | 14.8 | 35.7 | 39 | |
| VARD-TransMethod category=Non-LLM-based2026.03 | 10 | 14.8 | 35.7 | 39 | |
| SCORER+CBRMethod category=Non-LLM-based2026.03 | 10 | 15 | 33.4 | 39.6 | |
| VIXEN2024.08 | 8.6 | 15.4 | 38.1 | 42.5 | |
| VIXEN-CMethod category=LLM-based2026.03 | 8.6 | 15.4 | 38.1 | 42.5 | |
| CLIP4IDC2024.08 | 8.2 | 14.6 | 32.2 | 40.4 | |
| NCT2024.08 | 8.1 | 15 | 34.2 | 38.8 | |
| NCTMethod category=Non-LLM-based2026.03 | 8.1 | 15 | 34.2 | 38.8 | |
| BiDiff2024.08 | 6.9 | 14.6 | 27.7 | 38.5 | |
| DUDAMethod category=Non-LLM-based2026.03 | 6.5 | 12.4 | 22.8 | 37.3 |