Coarse component-wise spectral matching of pretrained GPT-2 weights changes structure but does not beat standard initialization, while direct weight reuse remains competitive.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Complexity-Guided Component-wise Initialization for Language Model Pretraining
Coarse component-wise spectral matching of pretrained GPT-2 weights changes structure but does not beat standard initialization, while direct weight reuse remains competitive.