{
  "blyg": "0.3",
  "id": "7zj9y401dwm5jvp1ws742tk86z",
  "kind": "fragment",
  "origin": "https://www.msweet.net/notes/",
  "page": "540-bronze-lynxes",
  "author": {
    "name": "Matthew McDowell-Sweet",
    "url": "https://www.msweet.net/"
  },
  "created": "2026-09-29T05:08:35Z",
  "updated": "2026-09-29T05:08:35Z",
  "version": 1,
  "content_md": "Hmmm.\n\n> But this phenomenon is not limited to weekdays. In the full paper, we find analogous structure in language models across more complex settings, including months, letters, ages, and a synthetic in-context learning task with predefined geometries. We also extend our results across modalities with an image-action model predicting the position of a car rolling up and down a hill (see the demo in our previous post). Across these tasks, manifold geometry provides a practical blueprint for steering model behavior.\n\n[Steering Along Manifolds to Control Neural Networks - Goodfire ↗](https://goodfire.com/research/manifold-steering#)",
  "content_html": "<p>Hmmm.</p>\n<blockquote><p>But this phenomenon is not limited to weekdays. In the full paper, we find analogous structure in language models across more complex settings, including months, letters, ages, and a synthetic in-context learning task with predefined geometries. We also extend our results across modalities with an image-action model predicting the position of a car rolling up and down a hill (see the demo in our previous post). Across these tasks, manifold geometry provides a practical blueprint for steering model behavior.</p></blockquote>\n<p class=\"source\"><a href=\"https://goodfire.com/research/manifold-steering#\">Steering Along Manifolds to Control Neural Networks - Goodfire ↗</a></p>",
  "content_hash": "sha256:751456c4179a7281c04112c2c6f512fef39595b75d73ced34d9a69a2fea6417e",
  "media": [],
  "changelog": [
    {
      "version": 1,
      "at": "2026-09-29T05:08:35Z",
      "note": null
    }
  ]
}