{
  "id": "prompt-optimization",
  "code": "PTL-0150",
  "term": "Prompt Optimization",
  "aliases": [
    "automatic prompt optimization",
    "automated prompt engineering"
  ],
  "category": "optimization",
  "definition": "Prompt optimization is the automatic search for a better prompt, in which a language model proposes candidate instructions or examples and each candidate is scored on an evaluation set until performance stops improving.",
  "description": "Methods differ in how they propose changes. Some resample paraphrases, as in APE, some show the optimizer a history of scored prompts, as in OPRO, some critique errors in natural language and edit accordingly, as in ProTeGi and TextGrad, and some use evolutionary search, as in Promptbreeder and EvoPrompt. All of them depend on a representative evaluation set and metric, and can overfit to it.",
  "example": null,
  "broader": [],
  "narrower": [
    "protegi",
    "textgrad",
    "promptbreeder",
    "evoprompt",
    "agentic-context-engineering"
  ],
  "related": [
    "automatic-prompt-engineer",
    "opro",
    "dspy",
    "evals"
  ],
  "introduced": null,
  "added": "2026-10-10",
  "sources": [
    {
      "title": "The Prompt Report: A Systematic Survey of Prompt Engineering Techniques",
      "authors": "Schulhoff et al.",
      "year": 2024,
      "url": "https://arxiv.org/abs/2406.06608"
    },
    {
      "title": "Automatic Prompt Optimization with \"Gradient Descent\" and Beam Search",
      "authors": "Pryzant et al.",
      "year": 2023,
      "url": "https://arxiv.org/abs/2305.03495"
    }
  ],
  "url": "https://protologue.com/t/prompt-optimization/",
  "citation": "Protologue. (2026). Prompt Optimization. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0150). https://protologue.com/t/prompt-optimization/"
}