method
active
method:prompt-label-baselinePrompt-Label Baseline
Conditional generation on explicit Big Five labels using per-dimension descriptors; used as inference-time baseline
Neighborhood — ranked by edge-count
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.
- Baseline model stitching trained in a single behavioral direction without CL auxiliary loss, used for comparison with CLMAS.
- A prompt framing requesting a representative sample from a distribution rather than a single instance, which is the key insight behind VS
- Baseline comparison method where models are directly prompted to be honest rather than fine-tuned
- Shows that explicit labels without latent steering fail to generalize to situational cues
- Baseline method sampling a random vector as feature direction for comparison with learned methods
- Control using objectively-NO factual questions under identical injection to measure global logit shift vs. genuine detection signal
- Testing five phrasings of the self-referential prompt to confirm robustness to wording variation
- Adaptation of instruction-tuned extraction to base models using third-person descriptions and hypothetical situations