What is the effect of model instructions on truth directions?

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

• Does instructing the model to assess correctness affect the geometry of truth directions?question0.847
  One of the three guiding research questions of the paper.
• We hypothesize that explicitly instructing the model to evaluate the correctness of the given statement may change the geometry of truth directions.hypothesis0.839
  Motivating hypothesis for Section 5's investigation of prompt template effects.
• Discovered truth directions are highly specific and do not interfere with general instruction-following behaviorclaim0.804
  Interpretation of KL divergence retention results
• Are the discovered truth directions robust to architectural variation and fine-tuning differences across model families?question0.801
  Open question on generalization beyond Gemma and Qwen families
• Universality claims for truth directions are more limited than previously assumed, with significant differences observable for various model layers, task difficulties, task types, and prompt templates.claim0.778
  Overarching conclusion summarizing the paper's contribution relative to prior universality claims.
• Will the no-prompt truth directions generalize to ask-correct activations?question0.776
  Specific question motivating the cross-template generalization experiment in Section 5.2.
• Truth Directionconcept0.775
  A hypothesized direction in LLM activation space that encodes the truth or falsehood of factual statements
• Multiple semantically adjacent truth directions make models more vulnerable to manipulations that shift outputs without obvious signs in the primary truth directionclaim0.773
  Safety implication derived from multi-dimensional truth structure finding