Injection stride s=1 produces the highest mean SJT scores across all LLMs; more frequent injection yields stronger steering

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

• MDS achieves global win proportion of 89.5% on SJTs across 14 LLMs and four injection stridesfinding0.789
  MDS dominates in open-ended generation by global win proportion metric (Table 2)
• MDS injection steering efficiency peaks at mid-layers across LLMs, injection strides, and OCEAN traitsfinding0.781
  Consistent empirical pattern supporting the connection between mid-layer representations and emotion/behavioral content
• PM achieves overall SJT steerability Phi=9.6 on gemma-3-12b-it vs MDS=8.7 and P2=8.3finding0.762
  Per-model steerability comparison from Table 4
• Random vectors at injection strength 8 elicit introspective awareness in 9 out of 100 trialsfinding0.754
  Random vectors are less effective, and even then produce introspection at lower rates.
• 47.69% of 130 injection-manipulated alpha trends have near-linear fits (R2 >= 0.95); 96.15% have roughly linear fits (R2 >= 0.75)finding0.750
  Demonstrates alignment with Linear Representation Hypothesis: target trait steers approximately linearly with alpha
• gemma-3-1b-it yields only one valid MDS injection score (phi_1,A,up = 4.8) and is excluded from main analysesfinding0.742
  Identified exception to overall MDS effectiveness; reason remains unexplained as a limitation
• Injection Stridemethod0.741
  Parameter controlling how often an injection is applied during completion; s=1 injects on every activation, achieving strongest steering
• MDS injections align with the Linear Representation Hypothesis: target trait varies near-linearly with alpha in open-ended generationclaim0.738
  Theoretical alignment claim backed by OLS R2 analysis showing 96.15% of trends have R2>=0.75