Position paper advocating that generative modeling over neural network weight spaces be standardized as a core primitive, based on the claim that high-performing models occupy structured low-dimensional regions.
Can we scale transformers to predict parameters of diverse ima- genet models? InICML, 2023
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Position: Weight Space Should Be a First-Class Generative AI Modality
Position paper advocating that generative modeling over neural network weight spaces be standardized as a core primitive, based on the claim that high-performing models occupy structured low-dimensional regions.