finding
active
finding:rank-1-tasks-k-32-2-show-empirical-transition-width-matching-n-crit-k-k-staircase-almost-perfectlyRank-1 tasks (K=32, β=2) show empirical transition width matching N_crit(k)=k staircase almost perfectly
Sharp test of utility ordering.
Source paper
extracted_from(2026) · Jing Huang · Daniel Wurgaft · Rachit Bansal · Laura Ruis +5
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Width buys features in order of utility, so only larger models learn infrequent or complex tasksassociated_withsupportsInterpretation of Theorem 3.
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.
- Fig. 3 and Fig. 19.
- E2 main interpretive claim.
- Interpretation of E2 results.
- Reconstruction cannot rank featurizers.
- Fig. 2 verification.
- Interpretation of abrupt behavior changes.
- Frontier ranking differs from other metrics.
- DAS achieves 100% IIA on hierarchical equality task with |N|=16, intervention size 8, Layer 1finding0.752DAS discovers a perfect alignment between the feed-forward network and the Both Equality Relations high-level model.