The ask-correct template delays truth direction emergence for F3 and reduces performance for F4-F5 compared to no-prompt.

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

• Truth directions fail to generalize to harder tasks (F3-F5) regardless of prompt template because activations remain highly entangled.claim0.839
  Establishes task difficulty as a hard limit that instructions cannot overcome.
• Using the ask-correct prompt improves cross-task generalization of arithmetic probes to factual tasks F0-F2.claim0.814
  Finding that explicit correctness framing partially aligns truth directions across task families.
• Truth probes fail to generalize to harder factual tasks F3-F5 regardless of prompt template, with AUROC near or below 0.6.finding0.796
  Establishes F3-F5 as a hard generalization boundary that instructions cannot overcome.
• Under ask-correct, probes trained on arithmetic tasks A1-A3 generalize almost perfectly to factual tasks F0-F2 (AUROC ~1.0), whereas under no-prompt this generalization is largely absent.finding0.784
  Key improvement in cross-task generalization enabled by explicit instruction framing.
• Will the no-prompt truth directions generalize to ask-correct activations?question0.783
  Specific question motivating the cross-template generalization experiment in Section 5.2.
• Pure factual-recall tasks F0-F2 show robust AUROC performance across all instruction template variations.claim0.770
  Contrasts with harder tasks that are sensitive to prompt variations.
• Does instructing the model to assess correctness affect the geometry of truth directions?question0.764
  One of the three guiding research questions of the paper.
• 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.claim0.760
  From the cross-task generalization heatmaps in Appendix B.3.3.