Local Linear Reconstruction Error

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

• reconstruction accuracyconcept0.738
  Metric of how well models reconstruct information from hidden states; Sauers' study found showing janus thread extends distribution tails.
• Linear Decodingmethod0.732
  Correlative technique measuring the type of information encoded in distributed representations via linear predictability.
• Linear Representation of Featuresconcept0.718
  The central object of study — the idea that a concept like truth is encoded as a direction in the LLM's latent space
• Linear representationconcept0.706
  The idea that features are encoded as directions in activation space.
• Algorithm 1: Finding Localist Alignment Matrixmethod0.703
  Algorithm that extracts a localist (axis-aligned) approximation from any learned orthogonal rotation matrix for baseline comparison.
• Linear Transformations (a :( b)concept0.702
• Local learning rulemethod0.702
  Learning rule where change in a parameter at point x,t depends only on system state at same or nearby spacetime points, without requiring global cost function computation
• Accuracy does not vary linearly with latent reflection directions; instead it follows a more non-linear mapping that requires deeper theoretical treatment.claim0.698
  Theoretical limitation identified by the authors distinguishing reflection from stylistic tasks.