claim
active
claim:no-single-l0-is-optimal-for-all-tasksNo single L0 is optimal for all tasks
Sparsity trade-off.
Source paper
extracted_from(2025) · Adam Karvonen · Can Rager · Johnny Lin · Curt Tigges +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Task-dependent optimal sparsity.
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.
- Argues against the single-layer analysis approach of prior work.
- Recommended strategy for gradient normalization.
- Caveat on metric.
- L1-of-norms is preferable for loss comparability and exposing layer- or model-specific featuresclaim0.749All experiments used L1-of-norms.
- Supports the claim against single-layer probing approaches used in prior work.
- Motivation for the proposed method.
- A combinatorial argument that good sequences are astronomically rare, emphasizing the difficulty of discovery.