Attention computations distribute across heads via parameter subcomponents with interpretable roles

{
  "doc": "bushnaq-goodfire-vpd-parameters-2026.md",
  "author": null,
  "context": "Mechanistic discovery about how attention mechanisms decompose into interpretable parameter components.",
  "enrichment": {
    "is_stale": true
  },
  "norm_label": "Attention computations distribute across heads via parameter subcomponents with interpretable roles",
  "source_url": null,
  "graphify_id": "finding_attention_distribution",
  "source_file": "bushnaq-goodfire-vpd-parameters-2026.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/bushnaq-goodfire-vpd-parameters-2026/graph.json",
  "extracted_type": "finding",
  "graphify_file_type": "finding"
}