Read As Human: Compressing Context via Parallelizable Close Reading and Skimming

Jiwei Tang, Shilei Liu, Zhicheng Zhang +9

Large Language Models (LLMs) demonstrate exceptional capability across diverse tasks. However, their deployment in long-context scenarios is hindered by two challenges: computation…