finding
active
finding:absorption-scores-worsen-with-dictionary-size-for-all-architectures-except-matryoshka-which-degrades-only-slightlyAbsorption scores worsen with dictionary size for all architectures except Matryoshka, which degrades only slightly
Inverse scaling.
Source paper
extracted_from(2025) · Adam Karvonen · Can Rager · Johnny Lin · Curt Tigges +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Matryoshka's hierarchical structure explains its positive scaling on disentanglement metricssupportsMechanistic interpretation.
Hypotheses (1)
hypothesis
- Explains inverse scaling.
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.
- Inverse scaling on SCR.
- Positive scaling.
- Frontier ranking differs from other metrics.
- Margin of advantage.
- Feature presence depends on concept frequency in training data, with a threshold scaling inversely with alive features.
- Central result.
- ReLU underperforms.
- Fig. 7d,e.