Long Context Alignment with Short Instructions and Synthesized Positions

Wenhao Wu, Yizhong Wang, Yao Fu +3

Effectively handling instructions with extremely long context remains a challenge for Large Language Models (LLMs), typically necessitating high-quality long data and substantial c…