MDS achieves global win proportion of 89.5% on SJTs across 14 LLMs and four injection strides

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 47.3% on MPI-120 inventory across 14 LLMsfinding0.883
  MDS is also the top method on the inventory task but with much smaller margin than on SJTs (Table 2)
• MDS injections outperform P2 in open-ended generation in 11 of 14 LLMs with Phi gains of 3.61% to 16.44%finding0.806
  Primary quantitative result overturning prior reports that prompting outperforms representation engineering
• Injection stride s=1 produces the highest mean SJT scores across all LLMs; more frequent injection yields stronger steeringfinding0.789
  Empirical finding about injection stride parameter; injecting into every completion activation maximizes steering strength
• PM hybrid outperforms both P2 and MDS in 13 of 14 LLMs with Phi gains over P2 from 5.56% to 21.92% and over MDS from 3.30% to 26.67%finding0.768
  Key finding showing that combining prompting and injection is the strongest approach
• On Qwen3-1.7B, MDS achieves ϕ1,C,↑ = 5.0 (SJTs) vs P2 at 4.7, and ϕ1,C,↓ = 1.4 (SJTs) vs P2 at 3.6finding0.765
  Specific consciousness sweep result for Qwen3-1.7B from Table 6 demonstrating strong bidirectional steering
• DAS achieves overall odds-ratio of 10.24 on pythia-410m averaged across all CausalGym tasksfinding0.751
  Numerical result for pythia-410m
• PM achieves overall SJT steerability Phi=9.6 on gemma-3-12b-it vs MDS=8.7 and P2=8.3finding0.750
  Per-model steerability comparison from Table 4
• MDS injections align with the Linear Representation Hypothesis: target trait varies near-linearly with alpha in open-ended generationclaim0.742
  Theoretical alignment claim backed by OLS R2 analysis showing 96.15% of trends have R2>=0.75