finding
active
finding:separating-hyperplane-in-orthogonal-characters-remaining-and-next-token-length-subspaces-achieves-auc-0-91-for-newline-predictionSeparating hyperplane in orthogonal characters-remaining and next-token-length subspaces achieves AUC 0.91 for newline prediction
Linear separability.
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Central thesis.
Questions (1)
question
- Intro question.
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.
- Causal validation.
- Ablating the character count subspace has a large loss effect only when the next token is a newlinefinding0.793Causal validation.
- Causal validation.
- Claim that geometry-to-behavior correlates exist
- Specificity of illusion.
- 271 stratified contexts.
- Table 3.
- Task competence.