question
active
question:do-psychological-steering-results-hold-beyond-64-token-completionsDo psychological steering results hold beyond 64-token completions?
Acknowledged limitation of restricting experiments to 64-token completions
Source paper
extracted_from(2026) · Leonardo Blas · Robin Jia · Emilio Ferrara
Neighborhood — ranked by edge-count
Papers (1)
paper
- Psychological Steering of Large Language Modelsassociated_with
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Shows that causal steering effects persist over long ranges for a substantial fraction of emotion probes
- Demonstrates long-tail persistence of causal steering effect in a subset of emotion features
- Justifies the choice of response tokens for persona vector extraction in the pipeline
- Demonstrates that the majority of emotion features show persistent upregulation shortly after a steering pulse
- At steering strength 1.4 all factual claims are incorrect yet the LLM evaluator still gives a high scorefinding0.765Pitfall of LLM-based steering evaluation.
- Shows immediate causal effect of steering on emotion feature activation
- Steering effect size across 35 emotion vectors correlates with probe-preference correlation at r=0.85finding0.753Emotions correlating with preference causally drive it.
- Demonstrates distributed steering is more effective and less accuracy-damaging than concentrated steering.