NLA-derived steering vectors from edited explanations can causally shift planning representations, changing rhyme completion from 'rabbit' to 'mouse' at ~50% success rate.

{
  "doc": "natural.md",
  "context": "Evidence that NLA explanations bear causal relationship to model outputs; demonstrates validity of extracted representations.",
  "category": "ai",
  "norm_label": "NLA-derived steering vectors from edited explanations can causally shift planning representations, changing rhyme completion from 'rabbit' to 'mouse' at ~50% success rate.",
  "graphify_id": "planning_steering_finding",
  "source_file": "natural.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/natural/graph.json",
  "extracted_type": "finding",
  "source_location": "§Planning in Poetry",
  "graphify_file_type": "finding"
}