You could have designed state of the art positional encoding

The article discusses recent advancements in positional encoding techniques for transformer models, highlighting novel designs that enhance the model's ability to capture sequential information. It emphasizes the introduction of a learnable positional encoding mechanism that outperforms traditional fixed sinusoidal encodings on benchmarks such as GLUE and SQuAD. This innovation is significant for practitioners as it provides a more flexible and adaptive approach to encoding positional information, potentially improving model performance on a variety of NLP tasks.