{
  "id": "activation-steering",
  "code": "PTL-0219",
  "term": "Activation Steering",
  "aliases": [
    "activation addition",
    "activation engineering",
    "steering vectors"
  ],
  "category": "optimization",
  "definition": "Activation steering is a technique for controlling a language model's behavior at inference time by adding a steering vector, typically the difference between the model's internal activations on contrasting inputs, to its hidden activations during the forward pass, without changing its weights.",
  "description": "A steering vector can be computed from as little as one pair of contrasting prompts, such as \"Love\" and \"Hate\", or by averaging activation differences across many positive and negative examples, a variant called contrastive activation addition. Scaling the vector up or down, or subtracting it, strengthens or weakens the targeted behavior or topic. It needs no gradient-based optimization, but it requires access to the model's internal activations, which hosted models do not expose.",
  "example": null,
  "broader": [],
  "narrower": [],
  "related": [
    "prompt-tuning",
    "low-rank-adaptation",
    "fine-tuning"
  ],
  "introduced": 2023,
  "added": "2026-10-20",
  "sources": [
    {
      "title": "Steering Language Models With Activation Engineering",
      "authors": "Turner et al.",
      "year": 2023,
      "url": "https://arxiv.org/abs/2308.10248"
    },
    {
      "title": "Steering Llama 2 via Contrastive Activation Addition",
      "authors": "Panickssery et al.",
      "year": 2023,
      "url": "https://arxiv.org/abs/2312.06681"
    }
  ],
  "url": "https://protologue.com/t/activation-steering/",
  "citation": "Protologue. (2026). Activation Steering. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0219). https://protologue.com/t/activation-steering/"
}