One-Token Likert Rating Extraction Protocol

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

• Response-Average Token Extractionmethod0.774
  Strategy of extracting persona vectors from averaged activations over response tokens, found most effective compared to prompt-based positions
• Response-average token extraction yields stronger steering effectiveness than prompt-last or prompt-average extraction positionsfinding0.708
  Justifies the choice of response tokens for persona vector extraction in the pipeline
• Single-Token Featuresconcept0.693
  Features that fire on every instance of a single token; appear in small dictionaries as collapsed versions of many token-in-context features
• Token-100 correlation persistence metricmethod0.693
  Measures emotion feature persistence as correlation between z-scored activation at token 0 and token 100 across all eligible target model tokens
• Prompt Token Approximation of Projection Differencemethod0.685
  Uses last prompt token projection to approximate base generation projection, avoiding expensive model rollouts
• whitening and z-scoring protocolmethod0.682
  Standardization of ρd, dr, and log k on dev set for computing S.
• PCA Analysis of Token Embeddings/Unembeddingsmethod0.680
  PCA applied to token embedding and unembedding matrices to understand what fraction of residual stream dimensions they occupy and how they relate
• Single-prompt concept vector extractionmethod0.678
  Method using activations from the prompt 'Tell me about {word}' minus mean over other random words to obtain concept vectors.