UNIFUSION: Adapting Autoregressive Language Models into Discrete Diffusion under a Unified Reverse-Rate Objective

• URL: http://arxiv.org/abs/2607.24507v1
• Date: Mon, 27 Jul 2026 14:40:26 GMT
• Title: UNIFUSION: Adapting Autoregressive Language Models into Discrete Diffusion under a Unified Reverse-Rate Objective
• Abstract summary: We propose a simple continual pre-training approach for adapting pretrained GPT2 checkpoints to uniform-noise diffusion.<n>ours steadily improves the trade-off between generative perplexity (GenPPL) and unigram entropy as the sampling budget increases.<n>At both scales, ours also achieves the highest WinoGrande, SIQA, and BBH accuracy among the compared diffusion models.
• Score: 17.48959208263029
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Existing methods mainly adapt pretrained autoregressive (AR) language models to masked diffusion, whereas we directly adapt them to uniform-noise diffusion, where every token remains editable during sampling. However, adapting AR checkpoints across corruption kernels remains challenging because existing DLMs use different objectives and prediction parameterizations. We establish connections among SEDD, MDLM/GIDD, M2S, and Neural CTMC by expressing their conditional losses as a single generalized Kullback--Leibler objective over model reverse rates. We further derive conversions from clean-token predictions to concrete-score, posterior-mean, and exit-rate/jump parameterizations, yielding a shared \(x_0\) interface that supports switching between mask and uniform kernels. Building on these connections, we propose \ours{}, a simple continual pre-training approach for directly adapting pretrained GPT2 checkpoints to uniform-noise diffusion. Through systematic evaluation of 124M- and 355M-parameter models, we show that \ours{} steadily improves the trade-off between generative perplexity (GenPPL) and unigram entropy as the sampling budget increases from 16 to 256 steps. At 256 steps, \ours{}-S and \ours{}-M achieve GenPPL/entropy pairs of \(97.783/5.2626\) and \(71.516/5.6669\), respectively; no evaluated model at the same scale simultaneously outperforms \ours{} on both metrics. At both scales, \ours{} also achieves the highest WinoGrande, SIQA, and BBH accuracy among the compared diffusion models.

Related papers

• Uniform Diffusion Models Revisited: Leave-One-Out Denoiser and Absorbing State Reformulation [72.05592785529404]
  We show that the standard plug-in bridge parameterization for UDM is not optimized by the denoising posterior.<n>We also introduce an absorbing-state reformulation of uniform diffusion that preserves the UDM joint law while decomposing it into masked-diffusion-like sampling operations.
  arXiv  Detail & Related papers  (2026-05-21T17:27:19Z)
• Posterior Augmented Flow Matching [64.1559809786948]
  Posterior-Augmented Flow Matching (PAFM) is a generalization of flow matching (FM)<n>We show that PAFM improves over FM by up to 3.4 FID50K across different model scales.
  arXiv  Detail & Related papers  (2026-05-01T17:59:59Z)
• Neural Continuous-Time Markov Chain: Discrete Diffusion via Decoupled Jump Timing and Direction [18.20305761007799]
  We propose textbfNeural CTMC, which parameterizes the reverse process through an emphexit rate (when to jump) and a emphjump distribution (where to jump) using two dedicated network heads.<n>We show that the evidence lower bound (ELBO) differs from a path-space KL divergence between the true and learned reverse processes by a $$-independent constant.<n>We also show that the tractable conditional surrogate preserves the gradients and minimizers of the corresponding marginal reverse-process objective under standard regularity assumptions.
  arXiv  Detail & Related papers  (2026-04-17T04:40:19Z)
• Self-Rewarding Sequential Monte Carlo for Masked Diffusion Language Models [58.946955321428845]
  This work presents self-rewarding sequential Monte Carlo (SMC)<n>Our algorithm stems from the observation that most existing MDLMs rely on a confidence-based sampling strategy.<n>We introduce the trajectory-level confidence as a self-rewarding signal for assigning particle importance weights.
  arXiv  Detail & Related papers  (2026-02-02T09:21:45Z)
• Inference-Time Scaling of Diffusion Language Models with Particle Gibbs Sampling [70.8832906871441]
  We study how to steer generation toward desired rewards without retraining the models.<n>Prior methods typically resample or filter within a single denoising trajectory, optimizing rewards step-by-step without trajectory-level refinement.<n>We introduce particle Gibbs sampling for diffusion language models (PG-DLM), a novel inference-time algorithm enabling trajectory-level refinement while preserving generation perplexity.
  arXiv  Detail & Related papers  (2025-07-11T08:00:47Z)
• Rectified Diffusion Guidance for Conditional Generation [94.83538269086613]
  We revisit the theory behind CFG and rigorously confirm that the improper combination coefficients (textiti.e.) brings about expectation shift the generative distribution.<n>We show that our approach enjoys a textbftextitform solution given the strength.<n> Empirical evidence on real-world data demonstrate the compatibility of our design with existing state-of-the-art diffusion models.
  arXiv  Detail & Related papers  (2024-10-24T13:41:32Z)
• Improving Vector-Quantized Image Modeling with Latent Consistency-Matching Diffusion [55.185588994883226]
  We introduce VQ-LCMD, a continuous-space latent diffusion framework within the embedding space that stabilizes training.<n>VQ-LCMD uses a novel training objective combining the joint embedding-diffusion variational lower bound with a consistency-matching (CM) loss.<n>Experiments show that the proposed VQ-LCMD yields superior results on FFHQ, LSUN Churches, and LSUN Bedrooms compared to discrete-state latent diffusion models.
  arXiv  Detail & Related papers  (2024-10-18T09:12:33Z)
• Model Inversion Attacks Through Target-Specific Conditional Diffusion Models [54.69008212790426]
  Model inversion attacks (MIAs) aim to reconstruct private images from a target classifier's training set, thereby raising privacy concerns in AI applications. Previous GAN-based MIAs tend to suffer from inferior generative fidelity due to GAN's inherent flaws and biased optimization within latent space. We propose Diffusion-based Model Inversion (Diff-MI) attacks to alleviate these issues.
  arXiv  Detail & Related papers  (2024-07-16T06:38:49Z)
• Efficient Integrators for Diffusion Generative Models [22.01769257075573]
  Diffusion models suffer from slow sample generation at inference time. We propose two complementary frameworks for accelerating sample generation in pre-trained models. We present a hybrid method that leads to the best-reported performance for diffusion models in augmented spaces.
  arXiv  Detail & Related papers  (2023-10-11T21:04:42Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.