Multi-modal Task Evaluation on MM-MergeBench Seen Tasks
83.74SciQA ScoreIndividual
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| IndividualTraining Mode=LoRA fine-tuned2026.03 | 83.74 | 96.02 | 67.58 | 43.4 | 65.5 | 64.8 | 57.29 | 75.54 | 69.23 | |
| Iso-CTSTraining Mode=LoRA fine-tuned2026.03 | 77.54 | 77.01 | 63.5 | 45.72 | 40.75 | 42.34 | 54.74 | 53.25 | 56.86 | |
| DC-MergeTraining Mode=LoRA fine-tuned2026.03 | 77.05 | 73.33 | 65.87 | 47.88 | 59.66 | 45.23 | 55.06 | 52.92 | 59.63 | |
| Multi TaskTraining Mode=LoRA fine-tuned2026.03 | 76.9 | 74.08 | 67.05 | 35.98 | 65.37 | 66.67 | 56.09 | 66.87 | 63.62 | |
| TSV-MTraining Mode=LoRA fine-tuned2026.03 | 75.66 | 78.61 | 59.48 | 41.92 | 41.71 | 41.19 | 52.23 | 45.97 | 54.6 | |
| RobustMergeTraining Mode=LoRA fine-tuned2026.03 | 73.43 | 65.54 | 67.2 | 44.8 | 62.97 | 46.61 | 52.8 | 45.9 | 57.33 | |
| Task ArithmeticTraining Mode=LoRA fine-tuned2026.03 | 71.94 | 57.49 | 67.06 | 38.9 | 62.87 | 44.8 | 49.2 | 39.21 | 53.93 | |
| DARE-MergingTraining Mode=LoRA fine-tuned2026.03 | 71.59 | 57.25 | 66.26 | 39.38 | 62.56 | 44.93 | 49.13 | 39.59 | 53.84 | |
| TIES-MergingTraining Mode=LoRA fine-tuned2026.03 | 71.49 | 55.88 | 66.73 | 39.67 | 65.12 | 44.35 | 47.06 | 34.46 | 53.09 | |
| PCB-MergingTraining Mode=LoRA fine-tuned2026.03 | 71.1 | 57.82 | 67.59 | 38.22 | 64.35 | 44.58 | 48.9 | 37.01 | 53.7 | |
| WUDI-MergingTraining Mode=LoRA fine-tuned2026.03 | 69.46 | 78.4 | 56.84 | 38.95 | 38.58 | 40.23 | 51.13 | 38.82 | 51.55 | |
| ZeroshotTraining Mode=LoRA fine-tuned2026.03 | 61.73 | 40.87 | 62.88 | 36.1 | 41.16 | 41.03 | 49.07 | 14.09 | 43.37 |