Drifting Objectives for Refining Discrete Diffusion Language Models

• URL: http://arxiv.org/abs/2605.19470v1
• Date: Tue, 19 May 2026 07:22:17 GMT
• Title: Drifting Objectives for Refining Discrete Diffusion Language Models
• Abstract summary: TokenDrift is a drifting objective that lifts categorical predictions to soft-token features.<n>It improves fixed-NFE generation quality over matched continuation baselines.<n>It reduces Gen.-PPL at 4 NFEs by 89% on MDLM and 86% on DUO.
• Score: 23.931675509910047
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Discrete diffusion language models (DDLMs) generate text by iteratively denoising categorical token sequences, while recent drifting methods for continuous generators suggest that part of this sampling-time correction can instead be absorbed into training through an anti-symmetric fixed-point objective. We study how to transfer this principle to DDLMs, where the main challenge is the interface with discrete text: hard token samples are non-differentiable, and categorical predictions do not directly provide continuous samples to drift. We formulate TokenDrift, a drifting objective that lifts categorical predictions to soft-token features, applies anti-symmetric drifting in a frozen semantic space, and backpropagates the resulting stop-gradient feature target to DDLM logits. In controlled continual-training experiments with masked and uniform-state diffusion backbones, TokenDrift improves fixed-NFE generation quality over matched continuation baselines, reducing Gen.-PPL at 4 NFEs by 89% on MDLM and 86% on DUO. These results suggest that drifting can provide a practical refinement objective for DDLMs.

Related papers

• Semantic DLM+: Improving Diffusion Language Models through Bias-variance Trade-off in Transition Kernel Design [59.05127237532803]
  Diffusion Language Models (DLMs) have demonstrated strong scaling capacity as alternatives to autoregressive language models.<n>We study this sensitivity through a principled analysis of generalization error and identify three critical factors.<n>We propose SemDLM+, which adds a global transition and a semantic-frequency penalty during sampling.
  arXiv  Detail & Related papers  (2026-06-13T14:41:55Z)
• 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)
• Sobolev Regularized MMD Gradient Flow [22.929984194375805]
  We propose a regularized variant of the gradient-regularized MMD flow.<n>The proposed regularization mitigates the difference between the mean kernel mean embeddings.<n>A key highlight of the proposed flow is that it is applicable in both sampling (from an unnormalized target distribution)
  arXiv  Detail & Related papers  (2026-05-12T09:58:19Z)
• Self-Distilled Trajectory-Aware Boltzmann Modeling: Bridging the Training-Inference Discrepancy in Diffusion Language Models [65.89572755202245]
  Diffusion Language Models (DLMs) offer stronger global awareness and highly parallel generation.<n>Post-training DLMs with standard Negative Evidence Lower Bound (NELBO)-based supervised fine-tuning remains inefficient.<n>We propose a self-distilled trajectory-based post-training framework that aligns training with the easy-to-hard structure of inference.
  arXiv  Detail & Related papers  (2026-05-12T09:39:06Z)
• 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)
• Training-Free Distribution Adaptation for Diffusion Models via Maximum Mean Discrepancy Guidance [17.353524034156205]
  MMD Guidance augments the reverse diffusion process with gradients of the Maximum Mean Discrepancy (MMD) between generated samples and a reference dataset.<n>Our framework naturally extends to prompt-aware adaptation in conditional generation models via product kernels.<n>Experiments on synthetic and real-world benchmarks demonstrate that MMD Guidance can achieve distributional alignment while preserving sample fidelity.
  arXiv  Detail & Related papers  (2026-01-13T09:42:57Z)
• Constrained Adaptive Rejection Sampling [27.579645342312674]
  Language Models (LMs) are increasingly used in applications where generated outputs must satisfy strict semantic or syntactic constraints.<n>Existing approaches to constrained generation fall along a spectrum: greedy constrained decoding methods enforce validity during decoding but distort the LM's distribution.<n>We present Constrained Adaptive Rejection Sampling (CARS), an approach that strictly improves the sample-efficiency of RS without distributional distortion.
  arXiv  Detail & Related papers  (2025-10-02T11:17:26Z)
• 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)
• Diffusion-PINN Sampler [6.656265182236135]
  We introduce a novel diffusion-based sampling algorithm that estimates the drift term by solving the governing partial differential equation of the log-density of the underlying SDE marginals via physics-informed neural networks (PINN) We prove that the error of log-density approximation can be controlled by the PINN residual loss, enabling us to establish convergence guarantees of DPS.
  arXiv  Detail & Related papers  (2024-10-20T09:02:16Z)
• Continual-MAE: Adaptive Distribution Masked Autoencoders for Continual Test-Time Adaptation [49.827306773992376]
  Continual Test-Time Adaptation (CTTA) is proposed to migrate a source pre-trained model to continually changing target distributions. Our proposed method attains state-of-the-art performance in both classification and segmentation CTTA tasks.
  arXiv  Detail & Related papers  (2023-12-19T15:34:52Z)

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.