SCA: Selective Compression Attention for Efficiently Extending the Context Window of Large Language Models.

Huanran Zheng, Wei Zhu, Xiaoling Wang

Browse the full EMNLP paper archive.