Diffusion language models exhibit path-dependent denoising because local conditionals are often incompatible, with global order gaps decomposing into sums of local circulations that serve as inference-only diagnostics for order invariance.
Journal of the Royal Statistical Society
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
-
Path-Dependent Denoising: A Non-Conservative Field Perspective on Order Collapse in Diffusion Language Models
Diffusion language models exhibit path-dependent denoising because local conditionals are often incompatible, with global order gaps decomposing into sums of local circulations that serve as inference-only diagnostics for order invariance.