claim
active
claim:multidimensional-counting-representations-enable-rotation-via-linear-maps-and-fine-resolution-without-norm-explosionMultidimensional counting representations enable rotation via linear maps and fine resolution without norm explosion
Role of extra dimensions.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (3)
finding
- Evidence for twisting.
- Haiku does not redirect probability mass from over-limit words to newline; distributions differ from newline-stripped promptassociated_withMechanism limitation.
- Multiple boundary heads have distinct overlapping response curves that tile characters remainingsupportsStereoscopic algorithm.
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.
- Learned rotations reveal that direct probes over standard activation bases would miss the actual causal role of representations.
- Core theoretical claim about the target of representation learning
- Author’s interpretive claim that the shared geometry is general and robust.
- Distinct from internal emotion.
- Mechanistic explanation for why superposition is geometrically feasible
- The idea that features are encoded as directions in activation space.
- Interpretive claim about what linear DAS results actually tell us
- Learning dynamics.