SEP 8, 2026 · PREPRINT
Let It Go or Learn to Self-Correct: Continuous Diffusion for Constrained Discrete Tasks
arXiv
This is a preprint presenting an algorithmic improvement to diffusion models on discrete constraint satisfaction tasks, supported by empirical results but lacking peer review and clinical or validated-domain applicability.
Reported
Sudoku validity (standard sampler)31%
Sudoku validity (direct clean pre…95%