How can mechanistic interpretability methods automatically identify attention computations that span multiple attention heads?

{
  "doc": "paper.md",
  "context": "Long-standing bottleneck in mechanistic interpretability that VPD addresses by working natively on attention weight matrices.",
  "category": "ai",
  "norm_label": "How can mechanistic interpretability methods automatically identify attention computations that span multiple attention heads?",
  "graphify_id": "attention_bottleneck_question",
  "source_file": "paper.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/paper/graph.json",
  "extracted_type": "question",
  "source_location": "§2.1",
  "graphify_file_type": "question"
}