Do psychological steering results hold beyond 64-token completions?

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

• 48 of 171 emotion probes individually significant at token 100 post-steeringfinding0.810
  Shows that causal steering effects persist over long ranges for a substantial fraction of emotion probes
• At 100 tokens post-steering, 48 of 171 emotion features remain individually BH-significant despite average effect being near zero.finding0.793
  Demonstrates long-tail persistence of causal steering effect in a subset of emotion features
• 62% of emotions significantly elevated at 5 tokens after steering pulse endsfinding0.768
  Demonstrates that the majority of emotion features show persistent upregulation shortly after a steering pulse
• At 5 tokens after steering pulse ends, 130 of 171 (62%) emotion features are BH-significantly elevated; 14% are suppressed.finding0.755
  Shows immediate causal effect of steering on emotion feature activation
• Steering at 6 layers (strength 0.6 each, total 3.6) outperforms single-layer steering at equivalent total strength for type hint suppressionfinding0.749
  Demonstrates distributed steering is more effective and less accuracy-damaging than concentrated steering.
• 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 layerclaim0.747
  Practical finding for optimizing steering setup.
• All-token steeringmethod0.745
  Baseline steering method that applies intervention at every token generation step, shown to degrade performance at high strengths
• 5-Token Steering Pulse Experimentmethod0.744
  Applies a 5-token steering pulse to each emotion probe and measures persistence of causal effect via contrast z-score over 200 subsequent tokens