Differentiable Parameter Propagation trains VLMs to emit design-parameter patches under hard layout and typography constraints, reaching 89% constraint satisfaction versus 52% for GPT-4V.
ACM SIGGRAPH 2025 Posters , series =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.GR 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
StructuredEdit: Constraint-Aware Graphic Design Editing via Differentiable Parameter Propagation
Differentiable Parameter Propagation trains VLMs to emit design-parameter patches under hard layout and typography constraints, reaching 89% constraint satisfaction versus 52% for GPT-4V.