B10 final accuracy 94.8 ± 1.2%

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

• B8 final accuracy 92.4 ± 1.8%finding0.887
  Accuracy at k=16 shots for B8.
• B9 final accuracy 89.7 ± 2.1%finding0.867
  Accuracy at k=16 shots for B9.
• Binary detection adjusted accuracy reaches 97.3% at layer 0 with α=5 before baseline control is appliedfinding0.791
  The misleadingly high result that prior paradigm would report as evidence of introspection
• B10 shot midpoint k50 = 0.28 ± 0.05 shots with accuracy 94.8 ± 1.2%finding0.763
  Lowest threshold condition in E2; near-zero/one-shot threshold consistent with high pretraining density
• Binary detection accuracy (up to 97.3% at L0 α=5) is entirely explained by global logit shifts (r=0.999 correlation with control)finding0.748
  Core negative result: the binary detection paradigm cannot distinguish genuine introspection from uniform output bias
• Strength comparison accuracy averages 47% at layers 15-30, indistinguishable from 50% chancefinding0.735
  Shows collapse of introspective capability at later layers in the strength comparison task
• B10 phase width Δk = 1.21 ± 0.18finding0.734
  Transition width (k90 – k10) for B10.
• Top-5 instructions by µ(1→2) at ℓ=12 achieve average cosine similarity .9893 and average accuracy .5645 on gsm8k_adv for Gemma3-4B-ITfinding0.733
  High cosine similarity for Gemma3 steering vectors suggests strong linear reflection structure.