NLA explanations grow more informative over training with FVE increasing from 0.3-0.4 to 0.6-0.8 roughly linearly in log(training steps)

{
  "doc": "natural.md",
  "context": "Quantitative evidence that NLA training produces increasingly informative explanations despite optimizing only for reconstruction.",
  "norm_label": "NLA explanations grow more informative over training with FVE increasing from 0.3-0.4 to 0.6-0.8 roughly linearly in log(training steps)",
  "graphify_id": "finding_fve_improvement",
  "source_file": "natural.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/natural/graph.json",
  "extracted_type": "finding",
  "source_location": "§Method, §Evaluating NLAs during training",
  "graphify_file_type": "finding"
}