← All notes

Summary

LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models (2023) compresses prompts so that inference is cheaper while still keeping the information the model actually uses. It is a practical answer to long, expensive contexts when you would rather shrink the prompt than grow the window.

What it introduced: a prompt-compression method that uses a smaller language model to drop or shorten tokens that contribute little, with evaluations of speed and task quality after compression.

What it led to: later prompt- and context-compression work in agent systems, including compaction and gateway accounting. Live notes: ao-cost-compression-accounting, ao-compaction-summarize-and-resume, and ao-context-long-context-limits.

Use it when you want to cut prompt tokens without training a new long-context model. Where to read it: https://arxiv.org/abs/2310.05736 (v2, last revised 6 Dec 2023). Accepted at EMNLP 2023.

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

@misc{iil_note_8mc7dsyh,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2310-05736-llmlingua/},
  note = {Permanent ID iil:note:8mc7dsyh. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2310-05736-llmlingua.md.