Long-sequence modeling on Long Range Arena (LRA) v1 (test)
63.14ListOpsMega
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Megacomplexity=O(L^2)2022.08 | 63.14 | 90.43 | 91.25 | 90.44 | 96.01 | 97.98 | 88.21 | |
| MegaComputational complexity=O(L^2)2022.08 | 63.14 | 90.43 | 91.25 | 90.44 | 96.01 | 97.98 | 88.21 | |
| Liquid-S42022.08 | 62.75 | 89.02 | 91.2 | 89.5 | 94.8 | 96.66 | 87.32 | |
| Liquid-S42022.08 | 62.75 | 89.02 | 91.2 | 89.5 | 94.8 | 96.66 | 87.32 | |
| S52022.08 | 62.15 | 89.31 | 91.4 | 88 | 95.33 | 98.58 | 87.46 | |
| S52022.08 | 62.15 | 89.31 | 91.4 | 88 | 95.33 | 98.58 | 87.46 | |
| DSS-Softmax2022.08 | 60.6 | 84.8 | 87.8 | 85.7 | 84.6 | 87.8 | 81.88 | |
| S4D-LinModel=S4D-Lin2022.06 | 60.52 | 86.97 | 90.96 | 87.93 | 93.96 | — | 78.39 | |
| S4D-Lin2022.08 | 60.52 | 86.97 | 90.96 | 87.93 | 93.96 | — | 78.39 | |
| S4D-LegSModel=S4D-LegS2022.06 | 60.47 | 86.18 | 89.46 | 88.19 | 93.06 | 91.95 | 84.89 | |
| S4D-LegS2022.08 | 60.47 | 86.18 | 89.46 | 88.19 | 93.06 | 91.95 | 84.89 | |
| S4D-LegS2022.08 | 60.47 | 86.18 | 89.46 | 88.19 | 93.06 | 91.95 | 84.89 | |
| S4-LegS/FouT2022.08 | 60.45 | 86.78 | 90.3 | 89 | 94.44 | — | 78.5 | |
| Converter2025.02 | 60.38 | 86.44 | 83.41 | 61.02 | 88.43 | — | 75.94 | |
| S4D-InvModel=S4D-Inv2022.06 | 60.18 | 87.34 | 91.09 | 87.83 | 93.78 | 92.8 | 85.5 | |
| S4D-Inv2022.08 | 60.18 | 87.34 | 91.09 | 87.83 | 93.78 | 92.8 | 85.5 | |
| S5-InvInitialization=Inverse2022.08 | 60.07 | 87.77 | 91.26 | 86.41 | 93.42 | 97.54 | 86.08 | |
| S5-LinInitialization=Linear2022.08 | 59.98 | 88.15 | 91.31 | 86.05 | 94.31 | 65.6 | 80.9 | |
| S4-LegSModel=S4-LegS2022.06 | 59.6 | 86.82 | 90.9 | 88.65 | 94.2 | 96.35 | 86.09 | |
| S4-LegS2022.08 | 59.6 | 86.82 | 90.9 | 88.65 | 94.2 | 96.35 | 86.09 | |
| S4-LegS2022.08 | 59.6 | 86.82 | 90.9 | 88.65 | 94.2 | 96.35 | 86.09 | |
| Mega-chunkcomplexity=O(L)2022.08 | 58.76 | 90.19 | 90.97 | 85.8 | 94.41 | 93.81 | 85.66 | |
| Mega-chunkComputational complexity=O(L)2022.08 | 58.76 | 90.19 | 90.97 | 85.8 | 94.41 | 93.81 | 85.66 | |
| S4 (original)Model=S4 (original)2022.06 | 58.35 | 76.02 | 87.09 | 87.26 | 86.05 | 88.1 | 80.48 | |
| S4-FouTModel=S4-FouT2022.06 | 57.88 | 86.34 | 89.66 | 89.07 | 94.46 | — | 77.9 | |
| S4-FouT2022.08 | 57.88 | 86.34 | 89.66 | 89.07 | 94.46 | — | 77.9 | |
| H-Transformer-1D2022.08 | 49.53 | 78.69 | 63.99 | 46.05 | 68.78 | — | 61.41 | |
| H-Trans.-1D2022.08 | 49.53 | 78.69 | 63.99 | 46.05 | 68.78 | — | 61.41 | |
| CCNN2022.08 | 43.6 | 84.08 | — | 88.9 | 91.51 | — | 68.02 | |
| CCNN2022.08 | 43.6 | 84.08 | — | 88.9 | 91.51 | — | 68.02 | |
| ParamixerVariation=Chord2025.02 | 39.71 | 78.87 | 78.73 | 44.68 | 79.16 | — | 58.91 | |
| Luna-256projection dimension=2562025.02 | 37.98 | 65.78 | 79.56 | 47.86 | 78.55 | — | 61.95 | |
| cosFormer2025.02 | 37.9 | 63.41 | 61.36 | 43.17 | 70.33 | — | 55.23 | |
| Reformer2022.05 | 37.27 | 56.1 | 53.4 | 38.07 | 68.5 | — | 50.67 | |
| Reformer2022.08 | 37.27 | 56.1 | 53.4 | 38.07 | 68.5 | — | 50.56 | |
| Reformer2025.02 | 37.27 | 56.1 | 53.4 | 38.07 | 68.5 | — | 50.67 | |
| Lunacontext length=2562022.08 | 37.25 | 64.57 | 79.29 | 47.38 | 77.72 | — | 59.37 | |
| Luna-2562022.08 | 37.25 | 64.57 | 79.29 | 47.38 | 77.72 | — | 59.37 | |
| Nyströmformer2022.08 | 37.15 | 65.52 | 79.56 | 41.58 | 70.94 | — | 57.46 | |
| Nyströmformer2025.02 | 37.15 | 65.52 | 79.56 | 41.58 | 70.94 | — | 58.95 | |
| Transformer + TLB2022.05 | 37.05 | 81.88 | 76.91 | 57.51 | 79.06 | — | 66.48 | |
| Synthesizer2022.05 | 36.99 | 61.68 | 54.67 | 41.61 | 69.45 | — | 52.88 | |
| Synthesizer2025.02 | 36.99 | 61.68 | 54.67 | 41.61 | 69.45 | — | 52.88 | |
| TransformerModel=Transformer2022.06 | 36.37 | 64.27 | 57.46 | 42.44 | 71.4 | — | 53.66 | |
| Transformer2022.08 | 36.37 | 64.27 | 57.46 | 42.44 | 71.4 | — | 53.66 | |
| Transformer2022.08 | 36.37 | 64.27 | 57.46 | 42.44 | 71.4 | — | 53.66 | |
| Vanilla Trans.2025.02 | 36.37 | 64.27 | 57.46 | 42.44 | 71.4 | — | 54.39 | |
| BigBird2022.05 | 36.05 | 64.02 | 59.29 | 40.83 | 74.87 | — | 55.01 | |
| BigBird2022.08 | 36.05 | 64.02 | 59.29 | 40.83 | 74.87 | — | 54.17 | |
| BigBird2025.02 | 36.05 | 64.02 | 59.29 | 40.83 | 74.87 | — | 55.01 | |
| Linformer2022.05 | 35.7 | 53.94 | 52.27 | 38.56 | 76.34 | — | 51.36 | |
| Linformer2025.02 | 35.7 | 53.94 | 52.27 | 38.56 | 76.34 | — | 51.36 | |
| Longformer2022.05 | 35.63 | 62.85 | 56.89 | 42.22 | 69.71 | — | 53.46 | |
| Longformer2025.02 | 35.63 | 62.85 | 56.89 | 42.22 | 69.71 | — | 53.46 | |
| FNet2022.08 | 35.33 | 65.11 | 59.61 | 38.67 | 77.8 | — | 54.42 | |
| FNet2025.02 | 35.33 | 65.11 | 59.61 | 38.67 | 77.8 | — | 55.3 | |
| Sinkhorn Transformer2022.05 | 33.67 | 61.2 | 53.83 | 41.23 | 67.45 | — | 51.29 | |
| Sinkhorn Trans.2025.02 | 33.67 | 61.2 | 53.83 | 41.23 | 67.45 | — | 51.29 | |
| Performer2022.05 | 18.01 | 65.4 | 53.82 | 42.77 | 77.05 | — | 51.41 | |
| Performer2022.08 | 18.01 | 65.4 | 53.82 | 42.77 | 77.05 | — | 51.18 | |
| Performer2025.02 | 18.01 | 65.4 | 53.82 | 42.77 | 77.05 | — | 51.41 | |
| Sparse Transformer2022.05 | 17.07 | 63.58 | 59.59 | 44.24 | 71.71 | — | 51.24 | |
| Sparse Trans.2025.02 | 17.07 | 63.58 | 59.59 | 44.24 | 71.71 | — | 51.24 | |
| Linear Transformer2022.05 | 16.13 | 65.9 | 53.09 | 42.34 | 75.3 | — | 50.55 | |
| Linear Trans.2022.08 | 16.13 | 65.9 | 53.09 | 42.34 | 75.3 | — | 50.46 | |
| Linear Trans.2025.02 | 16.13 | 65.9 | 53.09 | 42.34 | 75.3 | — | 50.55 |