Efficient Streaming Language Models with Attention Sinks.

Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, Mike Lewis

Browse the full ICLR paper archive.