finding
active
finding:vw-model-inflates-parameters-from-2-9m-to-331m-100VW model inflates parameters from 2.9M to ~331M (~100×)
Explicit pairwise weights.
Source paper
extracted_fromNicholas L. Turner · Jeffrey Wu · Joshua Batson
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- Rewriting of the 1L transformer into six families of virtual weights over tokens, positions, features, logits.
- Even restricted to most effective and helpful.
- Parameter count roughly matches original transformer.
- Empirical demonstration of VPD on a mid-scale transformer, establishing feasibility.
- Task balancing issue where surface normal prediction degrades under EW.
- Concurrent work result showing emergent misalignment occurs in small models
- Attributed to model anisotropy from saturation making hidden states harder to access
- Models produce first-attempt mean scores 87.8-91.8/100 without steering across all model familiesfinding0.698Establishes high baseline quality confirming steering-induced degradation is the experimental signal