{
  "title": "Attention weights mix a separate value branch",
  "caption": "Q and K give scaled scores. Masking and softmax give attention weights from these scores. V supplies the values on a separate branch. The author made this explanatory diagram to show data dependencies, not a kernel launch schedule. The playback is an example. Animation time is not measured runtime.",
  "alt": "Flow diagram: Q and K give masked attention weights. A separate V branch supplies the values for their weighted sum.",
  "sources": [
    {
      "path": "content/blog/2026-06-09_life-of-a-token-transformer/post.md",
      "sha256": "aca6d72d9a2cb565f5e2e514e33cc16c3f73abb24785570c2dd1a2ea672af45d",
      "description": "Current article explanation, equations, assumptions and retained citations."
    },
    {
      "path": "research/dynamic-mechanisms-2026-10-21/before/content/blog/2026-06-09_life-of-a-token-transformer/illustrations/02-attention-value-routing-algorithm.svg",
      "sha256": "9da59fcf5bea47b07565957c9942fb99bf9aff7ab9e88afad3d217408058e8eb",
      "description": "Frozen prior illustration; exact pre-migration bytes retained."
    },
    {
      "path": "content/blog/2026-06-09_life-of-a-token-transformer/illustrations/source-notes.json",
      "description": "Existing article illustration provenance, retained unchanged."
    }
  ],
  "limitations": [
    "The playback is an example. Playback duration is not measured runtime."
  ]
}