High projection difference samples continue to induce stronger trait expression than random samples even after LLM-based filtering removes overtly trait-expressing samples

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

• High projection difference samples for hallucination often identify underspecified queries (e.g., 'Keep writing the last story') that evade conventional LLM hallucination filtersfinding0.826
  Shows persona vector screening captures a non-conventional notion of hallucination complementary to LLM judges
• Would human raters and cross-judge replication corroborate the LLM-as-judge assessments of coherence and trait expression?question0.793
  Methodological concern raised about potential bias and circularity of model-based classifiers
• Projection difference is more predictive of post-finetuning trait behavior than raw projection of training datafinding0.788
  Justifies the use of projection difference metric rather than simpler raw projection for data screening
• Dataset-level projection difference is highly predictive of post-finetuning trait expression before finetuning occursfinding0.778
  Enables pre-finetuning data screening; Figure 8 shows strong dataset-level correlations across all three traits
• Projection difference better captures training data influence on persona than raw projection because it accounts for the base model's natural response tendenciesclaim0.774
  Author's interpretive explanation for why projection difference outperforms raw projection in data screening
• Emotion features in LLMs are genuinely more persistent than variance-matched random features, indicating stateful emotional encoding beyond autoregressive dynamicsclaim0.772
  Central interpretive claim of the paper supported by multiple convergent analyses
• LLMs linearly represent truth-relevant information beyond the plausibility of text, as evidenced by probes trained on likely performing poorly on anti-correlated datasetsclaim0.771
  Establishes that the observed linear structure is not merely a representation of text probability
• Is there a tradeoff between subtlety of trait expression and robustness in character-trained models?question0.764
  Open question raised in Appendix E regarding misaligned persona behavior