{
  "ok": true,
  "data": {
    "entry": {
      "id": "feature-probing",
      "kind": "method",
      "kindLabel": "Method",
      "title": "Probe for a concept with a linear classifier",
      "body": "Freeze the model, grab the hidden activations for a pile of examples, and train a tiny linear probe to predict some property (is this text about money? is it past tense?). If a simple line separates them, the concept is already represented in there, plain as day.",
      "tag": "interpretability",
      "date": "2025 · 09 · 30",
      "publishedAt": "2025-09-30",
      "note": "Cheap, fast, and weirdly revealing.",
      "insight": {
        "aha": "If a tiny linear classifier can read a concept from hidden states, the model has already organized that signal.",
        "tryThis": "Probe one concept across layers and look for where accuracy first rises above a simple baseline.",
        "watchFor": [
          "High probe accuracy mistaken for causal importance",
          "Labels that leak through surface wording",
          "Layer sweeps without a held-out validation split"
        ]
      },
      "extended": [
        "The setup is small: collect hidden states from a chosen layer for a labeled dataset, then fit logistic regression on top. High accuracy means the concept is linearly available at that layer.",
        "Sweep the probe across layers and you can watch where a concept first becomes readable. Abstract properties tend to sharpen in the middle layers, which is a clue about where the model is doing the relevant work."
      ],
      "url": "https://tips.platphormnews.com/entries/feature-probing"
    }
  }
}