claim
active
claim:narcissism-has-weaker-likelihood-signal-for-the-target-concept-giving-less-impact-of-icl-and-steeringNarcissism has weaker likelihood signal for the target concept, giving less impact of ICL and steering
Explanation of dataset difference.
Source paper
extracted_from(2025) · Eric Bigelow · Daniel Wurgaft · YingQiao Wang · Noah D. Goodman +3
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Dataset heterogeneity.
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.
- Steering operates by changing concept priors, while ICL accumulates evidence via likelihoodclaim0.771Mechanistic split.
- Mechanistic finding explaining why high-N personas are safe under steering
- Mechanistic ambiguity in interpreting why self-steering evaluation outperforms textual evaluation
- Validates using chain-of-thought belief monitoring as proxy for behavioral steering efficacy.
- Practical application proposed based on mechanistic findings
- Secondary mechanistic finding from trait-refusal alignment analysis
- Evidence of a bottleneck between richer internal variation and final report distribution in impulsivity→interest condition
- Prediction 3.