Pure factual-recall tasks F0-F2 show robust AUROC performance across all instruction template variations.

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

• Truth probes fail to generalize to harder factual tasks F3-F5 regardless of prompt template, with AUROC near or below 0.6.finding0.832
  Establishes F3-F5 as a hard generalization boundary that instructions cannot overcome.
• Factual tasks F0-F3 reach near-perfect AUROC in early-to-mid layers of Llama-3.1-8B; arithmetic tasks A1-A3 emerge much later; counting tasks F4-F5 emerge late similar to arithmetic.finding0.811
  Core empirical finding about layer-dependent truth direction emergence across task types.
• 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.finding0.803
  Key improvement in cross-task generalization enabled by explicit instruction framing.
• Within-family factual generalization (F0-F2) is consistently strong across all models and prompt settings.finding0.780
  Establishes a reliable baseline for factual truth direction universality within simple factual recall.
• F0-trained probes in layers 4-10 show inverted separation on F1 (AUROC ≈ 0), systematically misclassifying true statements as false.finding0.775
  Demonstrates that early-layer probes capture sentence polarity rather than truth.
• The ask-correct template delays truth direction emergence for F3 and reduces performance for F4-F5 compared to no-prompt.finding0.770
  Shows instruction effects extend to harder factual tasks.
• 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.finding0.769
  Generalization evidence that truth probes are not invariant to model instructions.
• 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.claim0.761
  From the cross-task generalization heatmaps in Appendix B.3.3.