The middle layer residual stream features are causally implicated in multi-step reasoning.

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

• Single-layer analyses can be misleading because early-layer truth directions may reflect surface features with limited cross-task generalization.claim0.778
  Methodological critique of prior work that fixed a single layer for truth probing.
• The residual stream has a deeply linear structure, enabling virtual weights and path expansion analysisclaim0.776
  Architectural observation enabling the entire mathematical framework; the residual stream is purely a sum of linear projections
• The model converges to a more stable truth direction in middle-to-late layers, as evidenced by increasing cosine similarity between layer-wise probes.claim0.764
  Supported by the geometric transition visible in cosine similarity heatmaps for F0-F3.
• We hypothesize earlier-layer interventions allow more downstream computation to process and potentially correct the perturbationhypothesis0.763
  Post-hoc explanation for why steering at layer 33 rather than layer 50 produced better ESR behavior in Llama-3.3-70B
• Truth directions emerge in earlier layers for factual tasks and later layers for arithmetic tasks.claim0.763
  Core empirical claim about the layer-dependence of truth direction emergence as a function of task type.
• Four features will be fundamental: (1) all processes rethought as morphogenetic; (2) sequences link into a net; (3) continuous evolution becomes widespread; (4) an ethical obligation to heal the land emerges.claim0.762
  Vision of the emerging paradigm shift in society.
• At layer 12 (the layer analyzed by Burger et al. 2024), tP and tG explain similar fractions of truth-related variance (~0.33 each).finding0.760
  Shows that Burger et al.'s layer choice corresponds to a transitional phase, not a universal property.
• Reflective reasoning requires late-stage integration of semantic and reasoning signals, hence reflection-related directions emerge more clearly in higher network layers.claim0.760
  Interpretive claim about the locus of reflection in transformer architecture.