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.

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

• 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.finding0.849
  Key improvement in cross-task generalization enabled by explicit instruction framing.
• Under ask-correct, arithmetic tasks A1-A2 show gradual AUROC increase peaking only in final layers, unlike the sharp transition under no-prompt.finding0.815
  Shows that explicit instructions delay the emergence of truth directions in arithmetic tasks.
• Truth probes fail to generalize to harder factual tasks F3-F5 regardless of prompt template, with AUROC near or below 0.6.finding0.810
  Establishes F3-F5 as a hard generalization boundary that instructions cannot overcome.
• F0-trained probes in layers 4-10 show inverted separation on F1 (AUROC ≈ 0), systematically misclassifying true statements as false.finding0.810
  Demonstrates that early-layer probes capture sentence polarity rather than truth.
• Probes trained on A1 degrade significantly when evaluated on A2 and more on A3; training on A2 achieves only AUROC ~0.65 on A3.finding0.789
  Shows rapid generalization decay for arithmetic truth directions with each additional operation.
• Activation probing detects final answer belief earlier in CoT than CoT monitor on both models, with especially pronounced gap on easy MMLU questionsfinding0.781
  Comparative finding establishing activation probing as superior to text-level monitoring for early belief detection
• Probes trained under different explicit instruction prompts (ask-correct, ask-t/f, ask-able, ask-arith) are highly aligned with each other in cosine similarity.finding0.776
  Shows the passive vs. active divide is more important than the specific wording of 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.claim0.771
  From the cross-task generalization heatmaps in Appendix B.3.3.