claim
active
claim:in-the-dallas-example-the-model-performs-genuine-two-step-reasoning-internally-coexisting-with-shortcut-reasoningIn the Dallas example the model performs genuine two-step reasoning internally, coexisting with shortcut reasoning.
Central introductory claim.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (2)
finding
- Swapping Texas features for California features (from Oakland prompt) makes Claude 3.5 Haiku output Sacramento.associated_withsupportsValidates intermediate Texas representation.
- Confirms graph-predicted effects.
Concepts (1)
concept
- Coherent feedforward loopanalogous_toBiological circuit motif matching Dallas→Texas→Austin plus shortcut.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Distinct from sandbagging view of awareness.
- Not enough to trust medical responses without experts.
- Practical implication.
- Load-bearing interpretive claim about the layer-specificity of Burger et al.'s finding.
- Limitation acknowledged regarding generalizability of the reflection identification method
- Supports the awareness hypothesis.
- Practical interpretive upshot connecting dynamics to model competence.
- Suggestive evidence for language-independent truth representation in LLMs