Bottom-up interpretability explains computation in the model's own terms rather than imposing top-down abstractions

{
  "doc": "paper.md",
  "context": "VPD is positioned as advancing a paradigm shift from top-down mechanistic interpretability (activation-based) to parameter-centric, data-driven discovery.",
  "category": "ai",
  "norm_label": "Bottom-up interpretability explains computation in the model's own terms rather than imposing top-down abstractions",
  "graphify_id": "bottom_up_interpretability",
  "source_file": "paper.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/paper/graph.json",
  "extracted_type": "claim",
  "source_location": "abstract",
  "graphify_file_type": "claim"
}