protologue

Decomposed Prompting

Also called DecomP, modular prompting.

Decomposed prompting breaks a complex task into simpler sub-tasks, each delegated to its own prompt, model call, or symbolic function, with a controller prompt that decides which handler to call next.

Description

Each sub-task handler can be optimized, given its own examples, or replaced with code, and handlers can call further decompositions recursively. Compared with a single chain-of-thought prompt, this modularity helps on tasks with steps that models do poorly in one pass, such as long-input processing or symbolic manipulation.

Sources

  1. Khot et al. (2022). Decomposed Prompting: A Modular Approach for Solving Complex Tasks.

Cite this entry

Protologue. (2026). Decomposed Prompting. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0114). https://protologue.com/t/decomposed-prompting/

BibTeX
@misc{protologue_decomposed_prompting,
  title = {Decomposed Prompting},
  author = {{Protologue}},
  year = {2026},
  howpublished = {Protologue: A Taxonomy of Prompting and LLM Techniques, v1.4.0},
  note = {Entry PTL-0114},
  url = {https://protologue.com/t/decomposed-prompting/}
}

Markdown JSON