Continual Learning on Task Stream S
93.7Average Accuracy (Caltech)LADA
Evaluation Results
| Method | Links | |||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LADABlack-Box=x, Params=9.0 M, Evaluation Protocol=Last Accuracy, Sample Count=16-shot2026.06 | 93.7 | — | — | — | — | — | — | — | — | — | — | — | 49.6 | 69.3 | 86.9 | 96.7 | 86.9 | 93.8 | 93.7 | 84.6 | 76 | 83.1 | — | — | — | — | — | — | — | — | — | — | — | |
| LADABlack-Box ?=×, Params=9.0 M2026.06 | 93.6 | — | 75.2 | 36.1 | 36.7 | 65.6 | 83.9 | 45.2 | 88 | 65.3 | 61.1 | 61.9 | 53.9 | 66.6 | 78 | 85.3 | 86.7 | 65.2 | 89.9 | 69.7 | 62.7 | 75.2 | 55.5 | 96.2 | 75.8 | 95.8 | 98.4 | 89.6 | 98.8 | 94.5 | 87.3 | 77.2 | 86.9 | |
| LADABlack-Box=x, Params=9.0 M, Evaluation Protocol=Average Accuracy, Sample Count=16-shot2026.06 | 91 | — | — | — | — | — | — | — | — | — | — | — | 49.1 | 61.3 | 71.6 | 84.4 | 85 | 62.8 | 89.7 | 69.2 | 62.9 | 72.7 | — | — | — | — | — | — | — | — | — | — | — | |
| Primal-RAILBlack-Box ?=√, Params=18.0 M2026.06 | 89.6 | — | 74.3 | 36.4 | 37.4 | 64.1 | 83.4 | 43.9 | 87.8 | 65.5 | 60.8 | 61.5 | 48.8 | 59 | 74.4 | 84 | 86.1 | 65.6 | 89.5 | 68.9 | 62.3 | 72.8 | 45.8 | 94.1 | 70.7 | 94.2 | 96.5 | 89 | 98.1 | 93.5 | 82 | 76.5 | 84 | |
| BETABlack-Box=√, Params=0.05 M, Evaluation Protocol=Last Accuracy, Sample Count=16-shot2026.06 | 88 | — | — | — | — | — | — | — | — | — | — | — | 46.7 | 71 | 88.4 | 96.9 | 86.2 | 92 | 92.8 | 83.7 | 75 | 82.1 | — | — | — | — | — | — | — | — | — | — | — | |
| BETABlack-Box ?=√, Params=0.05 M2026.06 | 86.2 | — | 76.8 | 34.9 | 47 | 70.8 | 76.4 | 43.1 | 89.2 | 62.8 | 65 | 62.9 | 49.4 | 66.2 | 81.2 | 87.4 | 83 | 64.3 | 90.5 | 67.6 | 66.2 | 74.2 | 54.6 | 92.2 | 75.8 | 95.9 | 98.5 | 89.7 | 96.1 | 93.8 | 86.8 | 77.1 | 86 | |
| BETABlack-Box=√, Params=0.05 M, Evaluation Protocol=Average Accuracy, Sample Count=16-shot2026.06 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | 46.5 | 62.5 | 75.8 | 86.2 | 80.7 | 62.8 | 90.2 | 66.9 | 65.6 | 72.2 | — | — | — | — | — | — | — | — | — | — | — | |
| LwFBlack-Box=x, Params=149.6 M, Evaluation Protocol=Last Accuracy, Sample Count=16-shot2026.06 | 83.1 | — | — | — | — | — | — | — | — | — | — | — | 20.9 | 47.5 | 38.2 | 75.5 | 84.7 | 50.1 | 78 | 75.8 | 74.6 | 62.8 | — | — | — | — | — | — | — | — | — | — | — | |
| LwFBlack-Box ?=×, Params=149.6 M2026.06 | 81.2 | — | 62.4 | 27.8 | 10.7 | 52 | 76 | 25.4 | 68.3 | 30.4 | 54.5 | 45.3 | 25.8 | 48.1 | 33.1 | 57 | 75.1 | 54.9 | 74.5 | 35.5 | 57.1 | 54.2 | 9.6 | 77.1 | 55.3 | 38.7 | 60.5 | 83.1 | 99.5 | 85.9 | 49.6 | 80 | 63.9 | |
| ZSCLBlack-Box ?=×, Params=149.6 M2026.06 | 80.8 | — | 71.3 | 33.8 | 33 | 66.2 | 85.2 | 40.2 | 81.9 | 57.3 | 62.5 | 59 | 39.7 | 52.9 | 40.8 | 79.3 | 88 | 51.4 | 85.5 | 62.9 | 64.1 | 64.5 | 33.8 | 80.4 | 60.2 | 31.1 | 85.8 | 91.3 | 80.4 | 93.7 | 84.9 | 79 | 72.1 | |
| WiSE-FTBlack-Box ?=×, Params=149.6 M2026.06 | 79.8 | — | 59.9 | 25.4 | 10.2 | 43.9 | 67.9 | 29.4 | 57.4 | 24.1 | 50.3 | 40.9 | 14.9 | 45.3 | 17.1 | 55.7 | 70.1 | 52 | 68 | 35.6 | 53.3 | 49.2 | 18.1 | 84.9 | 53.4 | 27 | 69.6 | 88 | 88.4 | 91.5 | 76.7 | 80.2 | 67.8 | |
| LwFBlack-Box=x, Params=149.6 M, Evaluation Protocol=Average Accuracy, Sample Count=16-shot2026.06 | 79.7 | — | — | — | — | — | — | — | — | — | — | — | 24.7 | 38.3 | 36.9 | 63.9 | 81 | 36.5 | 71.9 | 42.7 | 56.7 | 53.2 | — | — | — | — | — | — | — | — | — | — | — | |
| MoE-AdaptersBlack-Box ?=×, Params=59.8 M2026.06 | 79.4 | — | 69.5 | 30.8 | 19 | 60.3 | 85.6 | 43.7 | 85.6 | 55.5 | 57.3 | 56.4 | 52.4 | 57.7 | 42.7 | 81.1 | 86.6 | 64.8 | 86.7 | 61.3 | 59 | 67.2 | 51.9 | 79 | 64.2 | 51.5 | 95.1 | 87.6 | 96.4 | 89.1 | 84.4 | 74 | 77.5 | |
| MoE-AdaptersBlack-Box=x, Params=59.8 M, Evaluation Protocol=Last Accuracy, Sample Count=16-shot2026.06 | 78.7 | — | — | — | — | — | — | — | — | — | — | — | 43.2 | 57.6 | 32.8 | 79.4 | 86 | 86.7 | 87.8 | 78.2 | 74.2 | 70.5 | — | — | — | — | — | — | — | — | — | — | — | |
| MoE-AdaptersBlack-Box=x, Params=59.8 M, Evaluation Protocol=Average Accuracy, Sample Count=16-shot2026.06 | 77.9 | — | — | — | — | — | — | — | — | — | — | — | 43.6 | 52.1 | 34.7 | 75.9 | 86.3 | 45.2 | 87.4 | 66.6 | 60.2 | 63 | — | — | — | — | — | — | — | — | — | — | — | |
| WiSE-FTBlack-Box=x, Params=149.6 M, Evaluation Protocol=Last Accuracy, Sample Count=16-shot2026.06 | 76.8 | — | — | — | — | — | — | — | — | — | — | — | 21.8 | 42.9 | 20.8 | 77.5 | 84.9 | 30.7 | 76.6 | 75.8 | 72.5 | 58 | — | — | — | — | — | — | — | — | — | — | — | |
| WiSE-FTBlack-Box=x, Params=149.6 M, Evaluation Protocol=Average Accuracy, Sample Count=16-shot2026.06 | 76.5 | — | — | — | — | — | — | — | — | — | — | — | 27.1 | 40.9 | 31.3 | 68.7 | 81.6 | 31.4 | 74.7 | 51.7 | 58.4 | 54.2 | — | — | — | — | — | — | — | — | — | — | — | |
| BETABlack-Box=√, Params=0.05 M, Evaluation Protocol=Transfer Accuracy, Sample Count=16-shot2026.06 | 75.9 | — | — | — | — | — | — | — | — | — | — | — | — | 34.7 | 46.6 | 70.1 | 75.2 | 43.2 | 89.2 | 62.7 | 64.5 | 62.5 | — | — | — | — | — | — | — | — | — | — | — | |
| ZSCLBlack-Box=x, Params=149.6 M, Evaluation Protocol=Last Accuracy, Sample Count=16-shot2026.06 | 75.3 | — | — | — | — | — | — | — | — | — | — | — | 33.1 | 43.5 | 35.2 | 74.6 | 87.4 | 50.4 | 84.2 | 77.3 | 73.4 | 63.4 | — | — | — | — | — | — | — | — | — | — | — | |
| LADABlack-Box=x, Params=9.0 M, Evaluation Protocol=Transfer Accuracy, Sample Count=16-shot2026.06 | 75 | — | — | — | — | — | — | — | — | — | — | — | — | 36.1 | 35.9 | 66.3 | 83.7 | 42.1 | 88 | 65.3 | 61.4 | 61.5 | — | — | — | — | — | — | — | — | — | — | — | |
| ZSCLBlack-Box=x, Params=149.6 M, Evaluation Protocol=Average Accuracy, Sample Count=16-shot2026.06 | 75 | — | — | — | — | — | — | — | — | — | — | — | 36 | 40.7 | 40.5 | 71 | 85.3 | 46.3 | 83.3 | 60.7 | 61.5 | 60 | — | — | — | — | — | — | — | — | — | — | — | |
| Zero-shotSample Count=16-shot2026.06 | 74.3 | — | — | — | — | — | — | — | — | — | — | — | 23.8 | 36.4 | 37.4 | 64.1 | 83.4 | 43.9 | 87.8 | 65.5 | 60.8 | 57.7 | — | — | — | — | — | — | — | — | — | — | — | |
| ZSCLBlack-Box=x, Params=149.6 M, Evaluation Protocol=Transfer Accuracy, Sample Count=16-shot2026.06 | 73.3 | — | — | — | — | — | — | — | — | — | — | — | — | 32.6 | 36.8 | 62.1 | 83.8 | 42.1 | 83.6 | 56.5 | 60.2 | 59 | — | — | — | — | — | — | — | — | — | — | — | |
| MoE-AdaptersBlack-Box=x, Params=59.8 M, Evaluation Protocol=Transfer Accuracy, Sample Count=16-shot2026.06 | 71 | — | — | — | — | — | — | — | — | — | — | — | — | 34.9 | 19.2 | 63 | 86.6 | 20 | 87.2 | 63.7 | 58.6 | 56 | — | — | — | — | — | — | — | — | — | — | — | |
| WiSE-FTBlack-Box=x, Params=149.6 M, Evaluation Protocol=Transfer Accuracy, Sample Count=16-shot2026.06 | 70.1 | — | — | — | — | — | — | — | — | — | — | — | — | 31.9 | 25.3 | 56.3 | 79.8 | 29.9 | 74.9 | 45.6 | 56.8 | 52.3 | — | — | — | — | — | — | — | — | — | — | — | |
| LwFBlack-Box=x, Params=149.6 M, Evaluation Protocol=Transfer Accuracy, Sample Count=16-shot2026.06 | 66.6 | — | — | — | — | — | — | — | — | — | — | — | — | 26.9 | 19.5 | 51 | 78.4 | 26.6 | 68.9 | 35.5 | 56.1 | 47.7 | — | — | — | — | — | — | — | — | — | — | — | |
| Zero-shotBlack-Box ?=–, Params=–2026.06 | — | 23.8 | 74.3 | 36.4 | 37.4 | 64.1 | 83.4 | 43.9 | 87.8 | 65.5 | 60.8 | 57.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |