{
  "id": "promptbreeder",
  "code": "PTL-0153",
  "term": "Promptbreeder",
  "aliases": [],
  "category": "optimization",
  "definition": "Promptbreeder is an evolutionary prompt-optimization method in which a population of task prompts is mutated and selected by fitness on training examples, while the mutation prompts that generate the changes are themselves evolved.",
  "description": "Because the instructions for how to mutate prompts also improve over time, the method is described as self-referential. It discovered domain-specific prompts that outperformed hand-written chain-of-thought prompts on arithmetic and commonsense benchmarks, at the cost of many model calls.",
  "example": null,
  "broader": [
    "prompt-optimization"
  ],
  "narrower": [],
  "related": [
    "evoprompt",
    "opro",
    "automatic-prompt-engineer"
  ],
  "introduced": 2023,
  "added": "2026-10-10",
  "sources": [
    {
      "title": "Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution",
      "authors": "Fernando et al.",
      "year": 2023,
      "url": "https://arxiv.org/abs/2309.16797"
    }
  ],
  "url": "https://protologue.com/t/promptbreeder/",
  "citation": "Protologue. (2026). Promptbreeder. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0153). https://protologue.com/t/promptbreeder/"
}