Single-layer analyses can be misleading because early-layer truth directions may reflect surface features with limited cross-task generalization.

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

• No single layer is universally optimal for probing truth directions; different tasks peak at different layers.claim0.843
  Argues against the single-layer analysis approach of prior work.
• Truth directions emerge in earlier layers for factual tasks and later layers for arithmetic tasks.claim0.814
  Core empirical claim about the layer-dependence of truth direction emergence as a function of task type.
• The model converges to a more stable truth direction in middle-to-late layers, as evidenced by increasing cosine similarity between layer-wise probes.claim0.797
  Supported by the geometric transition visible in cosine similarity heatmaps for F0-F3.
• Layer-wise trajectories show early enrichment, mid-layer alignment, and late re-clustering.claim0.786
  Qualitative geometry pattern.
• Truth-related directions reliably emerge at 60–75% of normalized layer depth in Qwen and Gemma modelsfinding0.783
  Experiment 1 finding localizing where truth can be causally mediated
• For simple factual tasks F0-F3, probe directions show a sharp geometric transition in middle layers, with late-layer probes converging to high cosine similarity; A3 and F4-F5 show no clear transition.finding0.782
  Geometric evidence for convergence to stable truth directions only for simpler tasks.
• We hypothesize that explicitly instructing the model to evaluate the correctness of the given statement may change the geometry of truth directions.hypothesis0.779
  Motivating hypothesis for Section 5's investigation of prompt template effects.
• The middle layer residual stream features are causally implicated in multi-step reasoning.claim0.778
  Features for Kobe Bryant, California, Lakers participate in computing the capital answer.