LongAlign: A Recipe for Long Context Alignment of Large Language Models

Yushi Bai, Xin Lv, Jiajie Zhang +6

Extending large language models to effectively handle long contexts requires instruction fine-tuning on input sequences of similar length. To address this, we present LongAlign --…