Few-shot classification on Synthetic Few-shot (Avg.)
90.7ICL ScoreCoQE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoQEE=512, L=4, #Param=2.0M2026.03 | 90.7 | 86.7 | |
| CoQEE=256, L=4, #Param=1.0M2026.03 | 88.7 | 79.2 | |
| TransformerE=64, L=12, #Param=0.9M2026.03 | 64.4 | 47.3 | |
| Transformer + RegularizationE=64, L=12, #Param=0.9M2026.03 | 63.2 | 53.2 | |
| Transformer + NoiseE=64, L=12, #Param=0.9M2026.03 | 62 | 42.3 | |
| TransformerE=64, L=8, #Param=0.6M2026.03 | 59.9 | 50.1 | |
| TransformerE=256, L=8, #Param=1.2M2026.03 | 50.5 | 67.8 | |
| Transformer + Temporary ForgettingE=64, L=12, #Param=0.9M2026.03 | 50.5 | 39.7 | |
| Linear TransformerE=64, L=12, #Param=0.8M2026.03 | 50.5 | 49.8 | |
| TransformerE=256, L=12, #Param=1.4M2026.03 | 50.4 | 64.9 | |
| TransformerE=512, L=12, #Param=2.4M2026.03 | 50.4 | 82.4 | |
| TransformerE=256, L=4, #Param=1.0M2026.03 | 50.3 | 71.3 | |
| TransformerE=512, L=8, #Param=2.2M2026.03 | 50.3 | 85.5 | |
| TransformerE=512, L=4, #Param=2.0M2026.03 | 50.3 | 88.5 | |
| TransformerE=64, L=4, #Param=0.4M2026.03 | 50.2 | 54 |