Efficient Streaming Language Models with Attention Sinks - #AI
StreamingLLM: Efficient streaming technique enable infinite sequence lengths