Alternative tokenizations Yes/No vs yes/no vs true/false had no significant effect on steering outcomes or ASR

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

• Stepwise steering achieves over 5% accuracy improvement compared to all-token intervention at similar token budgetfinding0.774
  Key result demonstrating advantage of stepwise over all-token steering strategy
• Stepwise steering preserves accuracy while reducing cost, whereas all-token steering causes significant degradation at large intervention strengthsclaim0.773
  Comparative claim between the two steering strategies
• All-token steeringmethod0.749
  Baseline steering method that applies intervention at every token generation step, shown to degrade performance at high strengths
• Up to 33.6% reasoning tokens saved on MMLU subsets with stepwise steering while maintaining accuracy in larger modelsfinding0.736
  Maximum token savings achieved by ReflCtrl on non-mathematical general reasoning tasks
• Steering vectors discover effective triggers such as 'However' and 'Otherwise', consistent with prior reported reflection datasetsfinding0.733
  Validates that steering vectors capture reflection semantics by finding tokens reported in related work.
• Activation steering effectively biases latent representations but does not fully replicate the mechanisms triggered by explicit instruction.claim0.733
  Nuanced interpretive claim about the limits of steering as a mechanism for reflection enhancement.
• At 100 tokens post-steering, 48 of 171 emotion features remain individually BH-significant despite average effect being near zero.finding0.733
  Demonstrates long-tail persistence of causal steering effect in a subset of emotion features
• Steering's effect on verbalized chain-of-thought evaluation/deployment beliefs is a useful proxy for its effect on actual deployment behaviorclaim0.729
  Practical guidance for practitioners who lack ground-truth model organisms.