Meaning-preserving transformations (paraphrasing, translating to French, shuffling) cause only small drops in FVE.

{
  "doc": "natural.md",
  "context": "Evidence that NLAs do not encode hidden information in overt text structure; explanations are primarily semantic.",
  "category": "ai",
  "norm_label": "Meaning-preserving transformations (paraphrasing, translating to French, shuffling) cause only small drops in FVE.",
  "graphify_id": "steganography_test",
  "source_file": "natural.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/natural/graph.json",
  "extracted_type": "finding",
  "source_location": "§Measuring behavioral properties of NLAs",
  "graphify_file_type": "finding"
}