protologue

Chain of Density

Also called CoD summarization.

Chain of Density is a summarization prompt that has the model write a summary and then repeatedly rewrite it, adding a few missing important entities each time without making it longer, to produce progressively denser summaries.

Description

Each round forces compression and fusion of content, so later summaries carry more information per word. The authors found human readers preferred summaries partway through the chain, denser than a plain summary but not maximally dense, which makes the number of rounds a tunable trade-off between informativeness and readability.

Sources

  1. Adams et al. (2023). From Sparse to Dense: GPT-4 Summarization with Chain of Density Prompting.

Cite this entry

Protologue. (2026). Chain of Density. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0120). https://protologue.com/t/chain-of-density/

BibTeX
@misc{protologue_chain_of_density,
  title = {Chain of Density},
  author = {{Protologue}},
  year = {2026},
  howpublished = {Protologue: A Taxonomy of Prompting and LLM Techniques, v1.4.0},
  note = {Entry PTL-0120},
  url = {https://protologue.com/t/chain-of-density/}
}

Markdown JSON