LongScope. LongScope is a benchmark for assessing large language models' ability to handle fragmented knowledge in extended textual contexts.