finding
active
finding:pruning-least-effective-70-of-vw-weights-worsens-loss-by-0-01-natsPruning least effective 70% of VW weights worsens loss by ~0.01 nats
Fisher-effectiveness pruning at 30% density.
Source paper
extracted_fromNicholas L. Turner · Jeffrey Wu · Joshua Batson
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Related work.
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.
- 15% density.
- Parameter count roughly matches original transformer.
- Even restricted to most effective and helpful.
- Core finding demonstrating non-monotonic relationship between base capability and harness-benefit
- Limitation.
- Interpretation of effectiveness tail result.
- Pruning reduces 14-turn graph from 7,239 nodes and 10.4M edges to ~1,300 nodes and ~110K edgesfinding0.717Circuit-tracer pruning.
- Quantifies how much of the base model's diversity VS can recover compared to baseline prompting