Regular Language Recognition on Modular Arithmetic
100AccuracyRNN
Evaluation results
| Method | Links | |
|---|---|---|
| RNNBackbone=RNN2026.05 | 100 | |
| LSTMBackbone=LSTM2026.05 | 100 | |
| MLP-LDRUBackbone=MLP-LDRU2026.05 | 100 | |
| SD-SSMBackbone=SD-SSM2026.05 | 99.9 | |
| RegularGPTBackbone=RegularGPT2026.05 | 99.1 | |
| BBT-GRCBackbone=BBT-GRC2026.05 | 94 | |
| Gated DeltaNetBackbone=Gated DeltaNet2026.05 | 76.5 | |
| TF (NoPE)Backbone=Transformer, Positional Encoding=NoPE2026.05 | 76 | |
| TF (ALiBi)Backbone=Transformer, Positional Encoding=ALiBi2026.05 | 54.4 | |
| S5Backbone=S52026.05 | 47.6 | |
| TF (~RoPE)Backbone=Transformer, Positional Encoding=~RoPE2026.05 | 31.4 |