finding
active
finding:within-cell-standard-deviation-of-trait-expression-scores-mean-15-11-median-15-76-min-0-00-max-44-22-on-0-100-scale-grows-with-coefficientWithin-cell standard deviation of trait-expression scores: mean 15.11, median 15.76, min 0.00, max 44.22 on 0-100 scale; grows with coefficient
Documents heterogeneity of steered responses; steering increases response variance
Source paper
extracted_from(2026) · Winston Zeng · Ali Emami · J H Choi
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- Validates that internal evaluation set provides reliable proxy for broader behavioral tendencies
- Distribution-level finding showing polarization of trait preferences post character training
- Demonstrates robustness of the trichotomy classification to cutoff choice
- PC1 explains 82% of variance in factor analysis of 2224 data points across 6 scoring dimensionsfinding0.740Dimensions are not independent; composite score is the reliable signal; six dimensions useful for understanding how not how much
- Mean combined expression: N/N pairs 171.53, N/S pairs 118.93, S/S pairs 71.02 (on 0-200 scale)finding0.734Natural anchor presence raises the combined floor; absence of natural anchor permits S/S regime to collapse
- Models produce first-attempt mean scores 87.8-91.8/100 without steering across all model familiesfinding0.731Establishes high baseline quality confirming steering-induced degradation is the experimental signal
- An LLM-judge-assigned score from 0-100 indicating how strongly a model response exhibits a target personality trait
- Discriminant validity: composite scores are not reducible to verbosity