{
  "id": "in-context-learning",
  "code": "PTL-0017",
  "term": "In-Context Learning",
  "aliases": [
    "ICL"
  ],
  "category": "exemplars",
  "definition": "In-context learning (ICL) is a language model's ability to perform a task by conditioning on instructions or demonstrations in its prompt, without any update to its weights.",
  "description": "Identified as an emergent capability of large pretrained models in the GPT-3 paper, ICL underlies few-shot prompting. Studies show that models often rely more on the format and label space of demonstrations than on whether the demonstration labels are correct.",
  "example": null,
  "broader": [],
  "narrower": [
    "few-shot-prompting",
    "demonstration-label-sensitivity",
    "many-shot-in-context-learning"
  ],
  "related": [],
  "introduced": 2020,
  "sources": [
    {
      "title": "Language Models are Few-Shot Learners",
      "authors": "Brown et al.",
      "year": 2020,
      "url": "https://arxiv.org/abs/2005.14165"
    }
  ],
  "url": "https://protologue.com/t/in-context-learning/",
  "citation": "Protologue. (2026). In-Context Learning. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.0.0, PTL-0017). https://protologue.com/t/in-context-learning/"
}