claim
active
claim:sae-atoms-remain-one-dimensional-and-the-loss-rewards-reconstruction-without-rewarding-coherent-treatment-of-geometric-objectsSAE atoms remain one-dimensional and the loss rewards reconstruction without rewarding coherent treatment of geometric objects
Source of mismatch.
Source paper
extracted_from(2026) · Usha Bhalla · Thomas Fel · Can Rager · Sheridan Feucht +8
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Limitation statement.
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.
- Core critique of sparse autoencoders: they break the geometric structure of representations, making it harder to see the big picture.
- Open question in limitations.
- Limitation statement.
- Automated interpretability and specificity ratings show SAE features are clearer than MLP neurons.
- Proxy since superposition cannot be directly measured in real models.
- Manifold-level descriptions recover overarching semantic structure that SAE features miss.claim0.761Positive claim that geometric descriptions retain the conceptual coherence lost in atomized feature decompositions.
- Features act as local patch detectors.
- Surprising finding that the two evaluation methods diverge in their relationship with persistence