finding
active
finding:k-fac-edit-reduces-olmo-7b-dolma-strict-accuracy-from-99-9-to-3-4-bsn-6-0-svd-3-0K-FAC edit reduces OLMo-7B Dolma strict accuracy from 99.9% to 3.4% (BSN 6.0%, SVD 3.0%)
Table 1.
Source paper
extracted_from(2025) · Merullo, Jack · Vatsavaya, Srihita · Bushnaq, Lucius · Lewis, Owen
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (2)
claim
- K-FAC suppresses the broadest range of unseen memorized text without requiring labeled forget datasupportsPractical advantage over BSN.
- Top K-FAC eigenvectors correspond to shared generalizing structure, while low-curvature components underwrite recitationassociated_withBasis for editing.
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.
- Better transfer to unseen memorized data.
- Replication at smaller scale.
- Regularization effect.
- Disentanglement evidence.
- K-FAC edit on ViT-Base drops memorized train accuracy from 81.6% to 3.5%, GT accuracy to 66.5% (SVD 58.9%)finding0.796Table 2.
- Coherence comparison.
- BSN preserves token ranking slightly better.
- Logic sometimes improves.