notesum.ai
Published at October 18SurgeryV2: Bridging the Gap Between Model Merging and Multi-Task Learning with Deep Representation Surgery
cs.AI
Released Date: October 18, 2024
Authors: Enneng Yang, Li Shen, Zhenyi Wang, Guibing Guo, Xingwei Wang, Xiaocun Cao, Jie Zhang, Dacheng Tao

| Method | SUN397 | Cars | RESISC45 | EuroSAT | SVHN | GTSRB | MNIST | DTD | Avg. |
| Pretrained | 66.8 | 77.7 | 71.0 | 59.9 | 58.4 | 50.5 | 76.3 | 55.3 | 64.5 |
| Individual | 82.3 | 92.4 | 97.4 | 100 | 98.1 | 99.2 | 99.7 | 84.1 | 94.2 |
| Traditional MTL | 80.8 | 90.6 | 96.3 | 96.3 | 97.6 | 99.1 | 99.6 | 84.4 | 93.5 |
| Weight Averaging | 72.1 | 81.6 | 82.6 | 91.9 | 78.2 | 70.7 | 97.1 | 62.8 | 79.6 |
| Fisher Merging [19] | 69.2 | 88.6 | 87.5 | 93.5 | 80.6 | 74.8 | 93.3 | 70.0 | 82.2 |
| RegMean [24] | 73.3 | 81.8 | 86.1 | 97.0 | 88.0 | 84.2 | 98.5 | 60.8 | 83.7 |
| Task Arithmetic [36] | 73.9 | 82.1 | 86.6 | 94.1 | 87.9 | 86.7 | 98.9 | 65.6 | 84.5 |
| Ties-Merging [21] | 76.5 | 85.0 | 89.3 | 95.7 | 90.3 | 83.3 | 99.0 | 68.8 | 86.0 |
| AdaMerging [26] | 79.0 | 90.3 | 90.8 | 96.2 | 93.4 | 98.0 | 99.0 | 79.9 | 90.8 |
| Concrete AdaMerging [52] | 77.8 | 91.2 | 92.1 | 97.0 | 94.4 | 97.9 | 99.0 | 79.5 | 91.1 |
| Weight Averaging w/ Surgery (Ours) | 73.7 | 83.9 | 92.0 | 98.4 | 82.4 | 86.3 | 98.7 | 71.9 | 85.9 |
| Task Arithmetic w/ Surgery (Ours) | 75.7 | 84.4 | 93.1 | 98.8 | 91.3 | 93.4 | 99.1 | 76.1 | 89.0 |
| Ties-Merging w/ Surgery (Ours) | 76.5 | 85.9 | 93.7 | 99.2 | 89.7 | 92.0 | 99.1 | 78.1 | 89.3 |
| AdaMerging w/ Surgery (Ours) | 80.3 | 90.8 | 94.3 | 98.2 | 94.1 | 98.7 | 99.2 | 82.5 | 92.3 |
| Weight Averaging w/ Surgery‡ (Ours) | 75.6 | 84.7 | 95.6 | 99.1 | 86.3 | 97.3 | 99.2 | 81.4 | 89.9 |
| Task Arithmetic w/ Surgery‡ (Ours) | 76.7 | 85.4 | 95.7 | 99.4 | 92.7 | 98.5 | 99.2 | 81.0 | 91.1 |
| Ties-Merging w/ Surgery‡ (Ours) | 77.4 | 87.2 | 95.7 | 99.5 | 91.3 | 98.2 | 99.3 | 81.9 | 91.3 |
| AdaMerging w/ Surgery‡ (Ours) | 80.4 | 91.1 | 95.5 | 99.2 | 94.9 | 99.0 | 99.3 | 83.2 | 92.8 |
| Weight Averaging w/ SurgeryV2 (Ours) | 77.5 | 86.2 | 96.0 | 99.5 | 97.8 | 99.1 | 99.6 | 81.3 | 92.1 |
| Task Arithmetic w/ SurgeryV2 (Ours) | 80.4 | 88.0 | 96.6 | 99.6 | 97.8 | 99.1 | 99.6 | 81.7 | 92.8 |
| Ties-Merging w/ SurgeryV2 (Ours) | 80.0 | 88.7 | 96.3 | 99.7 | 97.9 | 99.0 | 99.6 | 81.5 | 92.8 |
| AdaMerging w/ SurgeryV2 (Ours) | 83.8 | 91.5 | 96.7 | 99.6 | 97.9 | 99.1 | 99.6 | 84.0 | 94.0 |