Successful RL agents exhibit causal emergence that predicts final reward early in training and aligns representational dynamics with reward improvement.

{
  "doc": "pigozzi-levin-causally-emergent-alignment-2026.md",
  "author": null,
  "context": "Central finding: causal emergence serves as a previously undisclosed axis of neural representation reorganization in learning agents.",
  "enrichment": {
    "is_stale": true
  },
  "norm_label": "Successful RL agents exhibit causal emergence that predicts final reward early in training and aligns representational dynamics with reward improvement.",
  "source_url": null,
  "graphify_id": "causally_emergent_alignment_hypothesis",
  "source_file": "pigozzi-levin-causally-emergent-alignment-2026.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/pigozzi-levin-causally-emergent-alignment-2026/graph.json",
  "extracted_type": "hypothesis",
  "graphify_file_type": "hypothesis"
}