Efficient Streaming Language Models with Attention Sinks – PaperGrep

https://papergrep.dev/paper/efficient-streaming-language-models-with-153ebe