Random word prefix prompts show emergence patterns similar to no-prompt, suggesting prompt length alone does not shift truth geometry.

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

• Random word prefix control prompt (random-prompt)method0.790
  Control prompt with random words of same length as ask-correct to isolate token-count confounds.
• The ask-arith prompt shows weaker generalization to factual tasks compared to other explicit prompts, suggesting a specialized arithmetic prompt does not create a unified truth direction across task families.claim0.774
  From the cross-task generalization heatmaps in Appendix B.3.3.
• With unrestricted vocabulary, models occasionally respond in non-English Yes/No equivalents (e.g., Sí, Nein) after truth-direction interventionsfinding0.770
  Suggestive evidence for language-independent truth representation in LLMs
• Calibrated few-shot prompting was a surprisingly weak baseline for truth classification compared to linear probesfinding0.748
  Unexpected finding that behavioral baseline underperforms representational probing approaches
• Will the no-prompt truth directions generalize to ask-correct activations?question0.747
  Specific question motivating the cross-template generalization experiment in Section 5.2.
• In Opus 4.1, the think word representation decays to baseline in the final layer because the strong next-token prediction drowns out other representationshypothesis0.747
  Explanation for the 'silent' thought phenomenon.
• No-prompt probes show significant AUROC performance drop when evaluated on ask-correct activations, especially at layers where arithmetic truth directions emerge under no-prompt.finding0.744
  Generalization evidence that truth probes are not invariant to model instructions.
• Practical context length limitations in language models lead to forgetting outside the window, constraining coherence over time.claim0.743
  Claim about engineering constraint reinforcing the theoretical no-order result