{
  "id": "min-p-sampling",
  "code": "PTL-0218",
  "term": "Min-p Sampling",
  "aliases": [
    "min_p",
    "min-p"
  ],
  "category": "foundations",
  "definition": "Min-p sampling is a truncation sampling method that keeps only tokens whose probability is at least a fixed fraction of the most likely token's probability, so that the cutoff tightens when the model is confident and loosens when it is not.",
  "description": "Unlike top-p sampling, which uses a fixed cumulative-probability mass, min-p scales its threshold by the top token's probability. It is offered as a sampling parameter in open-source inference engines such as vLLM and is usually combined with temperature. Its introducing paper reports better quality and diversity at high temperatures, but a later reanalysis argues that the evidence does not show min-p outperforming established sampling methods.",
  "example": null,
  "broader": [],
  "narrower": [],
  "related": [
    "top-p-sampling",
    "top-k-sampling",
    "temperature",
    "text-degeneration"
  ],
  "introduced": 2024,
  "added": "2026-10-20",
  "sources": [
    {
      "title": "Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs",
      "authors": "Nguyen et al.",
      "year": 2024,
      "url": "https://arxiv.org/abs/2407.01082"
    },
    {
      "title": "Min-p, Max Exaggeration: A Critical Analysis of Min-p Sampling in Language Models",
      "authors": "Schaeffer et al.",
      "year": 2025,
      "url": "https://arxiv.org/abs/2506.13681"
    },
    {
      "title": "Sampling Parameters",
      "authors": "vLLM",
      "year": 2024,
      "url": "https://docs.vllm.ai/en/v0.6.6/dev/sampling_params.html"
    }
  ],
  "url": "https://protologue.com/t/min-p-sampling/",
  "citation": "Protologue. (2026). Min-p Sampling. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0218). https://protologue.com/t/min-p-sampling/"
}