Mid-layers (6-15) achieve peak anchoring because semantic structure differentiates while maintaining coherence, forming a Goldilocks zone

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

• Layer-wise anchoring peaks in a 'Goldilocks zone' between early and late layers.claim0.858
  Qualitative characterization of optimal anchoring depth.
• Math and code tasks show strongest mid-layer anchoring on LLaMA (S ≈ −1.65 at layers 8-12)finding0.825
  Task-specific E3 finding showing compositional reasoning requires deeper processing
• Math and code tasks require deeper processing to bind complex patterns, as evidenced by strongest mid-layer anchoring at layers 8-12claim0.808
  Task-specific interpretation of E3 anchoring pattern differences
• In LLaMA-2-13B, salient linear structure in the top PCs rapidly emerges in early-middle layers, with this emergence occurring later for conjunctive statements than simple statementsfinding0.785
  Layer-wise emergence pattern supporting hierarchical development hypothesis
• Peak Layer-wise Anchoring Score (S_max)concept0.781
  Maximum of S(ℓ) across layers; geometry summary used to predict θ50
• Peak anchoring Sbmax and normalized area AUSN correlate with per-item success and internal shot midpoints θ50, providing a geometry-to-behavior bridge.claim0.770
  Main interpretation of E3.
• Layer-wise geometry summaries (Sbmax, AUSN) predict internal few-shot thresholds θ50claim0.766
  Claim that geometry-to-behavior correlates exist
• Thought detection peaks at ~2/3 layer depth; intention checking peaks at ~1/2 layer depth.finding0.765
  Lindsey (2026) differential layer performance explained by Janus's path combinatorics — different tasks use different path distributions.