Long-range sequence modeling on LRA 92 (test) benchmark leaderboard · SOTA2 Research

Long-range sequence modeling on LRA 92 (test)

63.14ListOps AccuracyMEGA

Evaluation Results

Method	Links
MEGA2023.11		63.14	90.43	91.25
Liquid S42023.11		62.75	89.02	91.2
Long Conv, Exp + Squash2023.11		62.2	89.6	91.3
S52023.11		62.15	89.31	91.4
SGConv2023.11		61.45	89.2	91.11
LRU2023.11		60.28	89.41	89.91
S4D-Inv2023.11		60.18	87.3	91.09
ChordMixer2023.11		60.12	88.82	90.17
S4-LegS2023.11		59.6	86.82	90.9
S4D2023.11		59.53	88.13	89.74
S4D-Inv2023.11		59.3	88.18	91.17
RIR-EBT-GRC2023.11		58.04	87.97	88.99
BBT-GRC2023.11		55.15	85.69	90.78
H-Transformer-1D2023.11		49.53	78.69	63.99
Reformer2023.11		37.27	56.1	53.4
Transformer2023.11		37.11	65.21	79.14
Transformer TLB2023.11		37.05	81.88	76.91
BigBird2023.11		36.05	64.02	59.29
Linformer2023.11		35.7	53.94	52.27
Longformer2023.11		35.63	62.85	56.89
Skinhorn Transformer2023.11		33.67	61.2	53.83
RIR-GRC2023.11		20.18	86.89	88.2
Performer2023.11		18.01	65.4	53.82
Sparse Transformer2023.11		17.07	63.58	59.59
Linear Transformer2023.11		16.13	65.9	53.09
Local Attention2023.11		15.82	52.98	53.39