Fetching the paper…

Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers · Around