Sequence-to-label Image Classification on Long Range Arena (LRA) Image (test)
90.6AccuracyMRConv-L
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MRConv-LArchitecture category=Convolution-based architectures2025.07 | 90.6 | 29.39 | 11.83 | |
| Liquid-S4Architecture category=RNN-based architectures2025.07 | 89.5 | 41.98 | 6.11 | |
| S4-LegSArchitecture category=RNN-based architectures, Bi-directional input processing=true2025.07 | 88.7 | 13.74 | 1.5 | |
| HGRNArchitecture category=RNN-based architectures, Bi-directional input processing=true, Positional encoding=true2025.07 | 88.7 | 78.63 | 23.83 | |
| S5Architecture category=RNN-based architectures, Bi-directional input processing=true2025.07 | 88 | 19.47 | 9 | |
| SGConvArchitecture category=Convolution-based architectures2025.07 | 88 | — | 11.83 | |
| S4Architecture category=RNN-based architectures2025.07 | 87.3 | 12.98 | 769.5 | |
| mGRADEArchitecture category=Our implementation2025.07 | 87.1 | 2.65 | 769.5 | |
| DSSSOFTMAXArchitecture category=RNN-based architectures2025.07 | 85.7 | 7.63 | 769.5 | |
| LRUArchitecture category=Our implementation2025.07 | 84.3 | 4.58 | 9 | |
| MambaArchitecture category=RNN-based architectures2025.07 | 70 | — | — |