finding
active
finding:diverse-data-mix-saes-had-far-more-dead-and-very-low-frequency-features-and-were-less-suitable-for-intervention-than-lmsys-saeDiverse data-mix SAEs had far more dead and very low-frequency features and were less suitable for intervention than LMSYS SAE
Qualitative and quantitative observation.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Hypotheses (2)
hypothesis
- Prediction following diversity-undertraining logic.
- Explains weaker datamix SAEs.
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.
- Claim that feature grounding enables interpretability metrics.
- Authors' experience-based claim.
- Limitations section.
- SAE latents depend on training data.
- Quantitative assessment of feature quality using clinical concepts across models.
- App. B.
- Geometric explanation for prior negative results.
- Core concern.