Vanilla diffusion models fail at compositional generation because score estimation error dominates and cannot be fixed by inference-time techniques when targets are out-of-distribution.
arXiv preprint arXiv:2505.13740 , year=
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
-
Catastrophic Compositional Generation: Why Vanilla Diffusion Models Fail to Extrapolate
Vanilla diffusion models fail at compositional generation because score estimation error dominates and cannot be fixed by inference-time techniques when targets are out-of-distribution.