Prelude and Coda Layers

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

• early layers (0–5)concept0.720
  Layers with weak anchoring due to generic representations.
• deeper layers (16–28)concept0.715
  Layers where anchoring weakens systematically due to representational drift.
• Math/code tasks S ≈ -1.65 at layers 8–12finding0.713
  Task-specific peak anchoring score for structured reasoning domains.
• Early layers of convolutional networks are more interchangeable than later layers across different architecturesfinding0.713
  Lenc & Vedaldi finding on layer-wise alignment
• Lindsey: Opus 4/4.1 show concept representations in middle layers that decay to baseline by final layer ('silent' internal process)finding0.708
  Cited to support enacted vs described reflection distinction; capable models show silent mid-layer processing
• Layer 24 (indexed at 8) of LLaMA3.1-8B on Hinting satisfies Criteria 1 and 2 under both IIT 3.0 and IIT 4.0 (temporal permutation).finding0.706
  One of the most promising cases; approximately corresponds to the 2/3 layer of LLaMA3.1-8B.
• Introspective signals appear in middle layers but are suppressed by later post-training-shaped layers.finding0.705
  Mechanistic finding by Lindsey (2026) explaining how contemplative prompt may work: enables mid-layer introspection to reach output.
• Approximations and prunings compose badly, so postpone them.claim0.702