Few-shot classification on Synthetic Few-shot (P_bursty=0.9, alpha=1)
92.1ICL ScoreCoQE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoQEE=512, L=4, #Param=2.0M2026.03 | 92.1 | 90 | |
| CoQEE=256, L=4, #Param=1.0M2026.03 | 90.3 | 82 | |
| TransformerE=64, L=12, #Param=0.9M2026.03 | 84.2 | 65.1 | |
| Transformer + RegularizationE=64, L=12, #Param=0.9M2026.03 | 80.5 | 64.9 | |
| Transformer + NoiseE=64, L=12, #Param=0.9M2026.03 | 80 | 60.4 | |
| TransformerE=64, L=8, #Param=0.6M2026.03 | 74.4 | 65.2 | |
| Linear TransformerE=64, L=12, #Param=0.8M2026.03 | 51.3 | 68.3 | |
| TransformerE=256, L=8, #Param=1.2M2026.03 | 50.8 | 73.1 | |
| TransformerE=512, L=12, #Param=2.4M2026.03 | 50.5 | 79.6 | |
| TransformerE=512, L=8, #Param=2.2M2026.03 | 50.4 | 81 | |
| Transformer + Temporary ForgettingE=64, L=12, #Param=0.9M2026.03 | 50.4 | 54.8 | |
| TransformerE=64, L=4, #Param=0.4M2026.03 | 50 | 66.2 | |
| TransformerE=256, L=4, #Param=1.0M2026.03 | 50 | 74.4 | |
| TransformerE=512, L=4, #Param=2.0M2026.03 | 49.9 | 82.5 | |
| TransformerE=256, L=12, #Param=1.4M2026.03 | 49.8 | 72.3 |