Gradient Dilution Issue

{
  "doc": "guo-atlas-2026-fulltext.md",
  "author": null,
  "context": "During RL training on ATLAS, sparse functional tokens (2.3% of sequences) receive diluted gradient signals from sequence-level advantages propagated across all tokens.",
  "category": "ai",
  "enrichment": {
    "is_stale": true
  },
  "norm_label": "Gradient Dilution Issue",
  "source_url": null,
  "graphify_id": "gradient_dilution",
  "source_file": "guo-atlas-2026-fulltext.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/guo-atlas-2026-fulltext/graph.json",
  "extracted_type": "finding",
  "source_location": "§2.3, Fig. 3",
  "graphify_file_type": "finding"
}