H1: Alignment training is attention training for models — Constitutional AI trains self-observation explicitly.

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Constitutional AI explicitly trains self-observation-like behavior, which is why CAI models score highest and show lowest contemplative lift.claim0.835
  Interpretive claim connecting the battery's circularity to the empirical finding
• H8: The contemplative system prompt provides external alignment equivalent to Constitutional AI training.hypothesis0.829
  Confirmatory hypothesis supported by calibrated lift data
• Constitutional AI can train a harmless but non-evasive AI assistant without any human harmfulness labels.claim0.816
  The paper's central claim, supported by findings that RL-CAI outperforms HH RLHF in harmlessness while being non-evasive.
• The contemplative system prompt provides externally what Constitutional AI alignment training provides internally.claim0.797
  Interpretation of the inverse relationship between CAI lift and default accessibility
• The constitutional approach makes it easier to control AI behavior precisely and with far fewer human labels.claim0.792
  Explicit principles replace large datasets of preference labels, enabling faster iteration.
• The only human oversight is provided through a list of rules or principles, and so we refer to the method as 'Constitutional AI'.quote0.789
  Defines the core concept of the paper.
• What predicts self-observation-like scores is training approach (alignment type), not model size or architecture.claim0.786
  Central interpretive claim from statistical analysis
• Reinforcement Learning Constitutional AIframework0.786
  The RL stage of CAI using AI feedback to train a preference model, then RL, resulting in a policy trained by RLAIF.