Distributing steering strength across multiple layers (6 layers at 0.6 each) is more effective and less accuracy-damaging than concentrating the same total strength in one layer

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

• Steering vector constructed from all 16 contrastive pairs outperforms most single-pair vectors; best 4-pair vector outperforms full 16-pair vectorfinding0.798
  Demonstrates averaging multiple prompt pairs reduces noise; optimal subset selection further improves performance.
• Inhibition steering produces larger accuracy drops than enhancement steering produces accuracy gains, across all models and datasets testedfinding0.779
  Key asymmetry finding: suppressing reflection is easier than inducing it.
• Performance is best when skipping both the first and last six layers when applying interventionclaim0.776
  Empirical configuration finding from ablation study on layer selection
• Steering vectors capture latent dimensions of reflective behavior more faithfully than surface-level embedding similarity.claim0.767
  Supported by the instruction discovery experiments comparing steering vs. embedding baselines.
• Stepwise steering preserves accuracy while reducing cost, whereas all-token steering causes significant degradation at large intervention strengthsclaim0.765
  Comparative claim between the two steering strategies
• Feature steering was effective in 5 out of 7 cases where few-shot probe steering vectors failed to produce meaningful behavior change.finding0.762
  Empirical comparison showing advantage of SAE features in low-data regime.
• Enhancement steering consistently underperforms compared to directly providing explicit reflection instructions across all tested conditionsfinding0.762
  Shows that activation steering does not fully replicate mechanisms triggered by explicit prompting.
• No single layer is universally optimal for probing truth directions; different tasks peak at different layers.claim0.759
  Argues against the single-layer analysis approach of prior work.