finding
active
finding:embedding-token-length-probes-1-14-show-a-circular-pattern-in-pc1-pc2-with-oscillating-pc3-3-pcs-capture-70-varianceEmbedding token-length probes (1-14) show a circular pattern in PC1-PC2 with oscillating PC3; 3 PCs capture 70% variance
Ringing at embedding.
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Toy model.
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.
- Supporting evidence for the claim that most residual stream dimensions are free for other layers to use
- Turn-averaged features have highest embedding similarity 0.663 at k=3; per-token max lowest at 0.617finding0.760Consistent across k=1,2,3,5.
- Core result of Experiment 3: cross-model semantic convergence under self-referential processing
- Intrinsic noise in representation.
- Evidence for twisting.
- Quantitative measure of emotion feature persistence vs random baseline in Cogito
- User and Assistant emotions dissociate.
- A random head in the same layer shows almost no structure between count and width probesfinding0.744Control for twisting.