Multi-domain Task Incremental Learning on MTIL benchmark Transfer 1.0 (test)
95.1Caltech101 AccuracyIndividual FT
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Individual FTFine-tuning Strategy=FT2025.03 | 95.1 | 62 | 89.6 | 79.5 | 98.9 | 97.5 | 92.7 | 99.6 | 94.7 | 89.6 | 81.8 | 89.2 | |
| Zero-shotFine-tuning Strategy=None2025.03 | 88.4 | 24.3 | 68.2 | 44.6 | 54.9 | 71 | 88.5 | 59.4 | 89 | 64.7 | 65.2 | 65.3 | |
| Dual-RAILTraining Paradigm=Transfer2025.03 | 88.4 | — | 68.2 | 44.6 | 54.9 | 71 | 88.5 | 59.6 | 89 | 64.7 | 65.2 | 69.4 | |
| DPeCLIPTraining Paradigm=Transfer2025.03 | 88.2 | — | 67.2 | 44.7 | 54 | 70.6 | 88.2 | 59.5 | 89 | 64.7 | 64.8 | 69.1 | |
| ConDUFine-tuning Strategy=LoRA, Training Paradigm=Transfer2025.03 | 88.1 | — | 68.9 | 45.7 | 57 | 71.3 | 88.8 | 61.2 | 89.3 | 65.1 | 67.8 | 70.3 | |
| ConDUFine-tuning Strategy=FT, Training Paradigm=Transfer2025.03 | 88.1 | — | 68.9 | 46.4 | 57.1 | 71.4 | 88.7 | 65.5 | 89.3 | 65 | 67.8 | 70.8 | |
| MulKITraining Paradigm=Transfer2025.03 | 87.8 | — | 69 | 46.7 | 51.8 | 71.3 | 88.3 | 64.7 | 89.7 | 63.4 | 68.1 | 70.1 | |
| ZSCLTraining Paradigm=Transfer2025.03 | 86 | — | 67.4 | 45.4 | 50.4 | 69.1 | 87.6 | 61.8 | 86.8 | 60.1 | 66.8 | 68.1 |