Compress Prompt
Compress prompt research focuses on reducing the length of input prompts for large language models (LLMs) to improve efficiency and reduce computational costs without sacrificing performance. Current approaches involve methods that selectively extract key information from prompts using techniques like parse tree analysis and graph-based representations, or that compress prompts into dense vectors using the LLM itself or other lightweight models. These advancements are significant because they address the limitations of long prompts in LLMs, enabling faster and more cost-effective processing of large datasets for various natural language processing tasks.
Papers
September 23, 2024
May 27, 2024
March 30, 2024
March 26, 2024
February 28, 2024