Classification on LRA Pathfinder N=1024 (test)
95.15AccuracyFull model
Evaluation Results
| Method | Links | |
|---|---|---|
| Full modelModel Backbone=S5, Pruning Ratio=0%, State Dimension=2562024.11 | 95.15 | |
| LASTModel Backbone=S5, Pruning Ratio=30%, State Dimension=1792024.11 | 94.45 | |
| Full modelModel Backbone=S4D, Pruning Ratio=0%, State Dimension=642024.11 | 87.94 | |
| LASTModel Backbone=S4D, Pruning Ratio=10%, State Dimension=582024.11 | 87.83 | |
| Uniform H_∞Model Backbone=S4D, Pruning Ratio=10%, State Dimension=582024.11 | 87.59 | |
| Global H_∞Model Backbone=S4D, Pruning Ratio=10%, State Dimension=582024.11 | 87.2 | |
| Paramixer (CHORD)Sequence Length=10242022.04 | 80.49 | |
| PerformerReference Source=[24], Sequence Length=10242022.04 | 77.05 | |
| LinformerReference Source=[24], Sequence Length=10242022.04 | 76.34 | |
| Linear TransformerReference Source=[24], Sequence Length=10242022.04 | 75.3 | |
| BigBirdReference Source=[24], Sequence Length=10242022.04 | 74.87 | |
| TransformerReference Source=[15], Sequence Length=10242022.04 | 74.16 | |
| Sparse TransformerReference Source=[24], Sequence Length=10242022.04 | 71.71 | |
| TransformerReference Source=[24], Sequence Length=10242022.04 | 71.4 | |
| Uniform H_∞Model Backbone=S5, Pruning Ratio=30%, State Dimension=1792024.11 | 71.38 | |
| NyströmformerReference Source=[15], Sequence Length=10242022.04 | 70.94 | |
| PerformerReference Source=[15], Sequence Length=10242022.04 | 69.87 | |
| LongformerReference Source=[24], Sequence Length=10242022.04 | 69.71 | |
| ReformerReference Source=[15], Sequence Length=10242022.04 | 69.36 | |
| ReformerReference Source=[24], Sequence Length=10242022.04 | 68.5 | |
| LinformerReference Source=[15], Sequence Length=10242022.04 | 67.6 | |
| Paramixer (CDIL)Sequence Length=10242022.04 | 67.13 | |
| Global H_∞Model Backbone=S5, Pruning Ratio=30%, State Dimension=1792024.11 | 57.2 |