Papers by JiqunChu JiqunChu
Incorporating Exponential Smoothing into MLP: a Simple but Effective Sequence Model (2024.findings-naacl)
Copied to clipboard
| Challenge: | Structured State Space models (SSMs) have been used for long-range sequence learning but are limited in their complexity and computational and memory requirements. |
| Approach: | They propose to incorporate a simple SSM into an element-wise MLP to reduce inductive bias. |
| Outcome: | The proposed model achieves comparable results to existing models on the LRA benchmark. |