English

Generalized Discrete Diffusion with Self-Correction

Machine Learning 2026-03-04 v1 Artificial Intelligence

Abstract

Self-correction is an effective technique for maintaining parallel sampling in discrete diffusion models with minimal performance degradation. Prior work has explored self-correction at inference time or during post-training; however, such approaches often suffer from limited generalization and may impair reasoning performance. GIDD pioneers pretraining-based self-correction via a multi-step BERT-style uniform-absorbing objective. However, GIDD relies on a continuous interpolation-based pipeline with opaque interactions between uniform transitions and absorbing masks, which complicates hyperparameter tuning and hinders practical performance. In this work, we propose a Self-Correcting Discrete Diffusion (SCDD) model to reformulate pretrained self-correction with explicit state transitions and learn directly in discrete time. Our framework also simplifies the training noise schedule, eliminates a redundant remasking step, and relies exclusively on uniform transitions to learn self-correction. Experiments at the GPT-2 scale demonstrate that our method enables more efficient parallel decoding while preserving generation quality.

Keywords

Cite

@article{arxiv.2603.02230,
  title  = {Generalized Discrete Diffusion with Self-Correction},
  author = {Linxuan Wang and Ziyi Wang and Yikun Bai and Wei Deng and Guang Lin and Qifan Song},
  journal= {arXiv preprint arXiv:2603.02230},
  year   = {2026}
}

Comments

40 pages, 3 figures, 6 tables