DeRaDiff: Denoising Time Realignment of Diffusion Models

• URL: http://arxiv.org/abs/2601.20198v1
• Date: Wed, 28 Jan 2026 02:53:39 GMT
• Title: DeRaDiff: Denoising Time Realignment of Diffusion Models
• Authors: Ratnavibusena Don Shahain Manujith, Yang Zhang, Teoh Tze Tzun, Kenji Kawaguchi,
• Abstract summary: How does one choose the right regularization strength?<n>DeRaDiff modulates the regularization strength during sampling to emulate models trained at other regularization strengths without any additional training or finetuning.<n>Our method consistently provides a strong approximation for models aligned entirely from scratch at different regularization strengths.
• Score: 45.23864679259705
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent advances align diffusion models with human preferences to increase aesthetic appeal and mitigate artifacts and biases. Such methods aim to maximize a conditional output distribution aligned with higher rewards whilst not drifting far from a pretrained prior. This is commonly enforced by KL (Kullback Leibler) regularization. As such, a central issue still remains: how does one choose the right regularization strength? Too high of a strength leads to limited alignment and too low of a strength leads to "reward hacking". This renders the task of choosing the correct regularization strength highly non-trivial. Existing approaches sweep over this hyperparameter by aligning a pretrained model at multiple regularization strengths and then choose the best strength. Unfortunately, this is prohibitively expensive. We introduce DeRaDiff, a denoising time realignment procedure that, after aligning a pretrained model once, modulates the regularization strength during sampling to emulate models trained at other regularization strengths without any additional training or finetuning. Extending decoding-time realignment from language to diffusion models, DeRaDiff operates over iterative predictions of continuous latents by replacing the reverse step reference distribution by a geometric mixture of an aligned and reference posterior, thus giving rise to a closed form update under common schedulers and a single tunable parameter, lambda, for on the fly control. Our experiments show that across multiple text image alignment and image-quality metrics, our method consistently provides a strong approximation for models aligned entirely from scratch at different regularization strengths. Thus, our method yields an efficient way to search for the optimal strength, eliminating the need for expensive alignment sweeps and thereby substantially reducing computational costs.

Related papers

• Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching [66.39914384073145]
  We propose a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates.<n>We find that step-level recombination is most beneficial on harder problems.<n>Our training-free framework improves average accuracy by up to 2 across six math and coding tasks.
  arXiv  Detail & Related papers  (2026-02-26T11:08:39Z)
• OSCAR: Orthogonal Stochastic Control for Alignment-Respecting Diversity in Flow Matching [14.664226708184676]
  Flow-based text-to-image models follow deterministic trajectories, forcing users to repeatedly sample to discover diverse modes.<n>We present a training-free, inference-time control mechanism that makes the flow itself diversity-aware.
  arXiv  Detail & Related papers  (2025-10-10T07:07:19Z)
• ADT: Tuning Diffusion Models with Adversarial Supervision [16.974169058917443]
  Diffusion models have achieved outstanding image generation by reversing a forward noising process to approximate true data distributions.<n>We propose Adrial Diffusion Tuning (ADT) to stimulate the inference process during optimization and align the final outputs with training data.<n>ADT features a siamese-network discriminator with a fixed pre-trained backbone and lightweight trainable parameters.
  arXiv  Detail & Related papers  (2025-04-15T17:37:50Z)
• Efficient Diffusion Training through Parallelization with Truncated Karhunen-Loève Expansion [5.770347328961063]
  Diffusion denoising models suffer from slow convergence during training.<n>We propose a novel forward-time process for training and sampling.<n>Our method significantly outperforms baseline diffusion models.
  arXiv  Detail & Related papers  (2025-03-22T05:34:02Z)
• Distributional Diffusion Models with Scoring Rules [83.38210785728994]
  Diffusion models generate high-quality synthetic data.<n> generating high-quality outputs requires many discretization steps.<n>We propose to accomplish sample generation by learning the posterior em distribution of clean data samples.
  arXiv  Detail & Related papers  (2025-02-04T16:59:03Z)
• Fast constrained sampling in pre-trained diffusion models [80.99262780028015]
  We propose an algorithm that enables fast, high-quality generation under arbitrary constraints.<n>Our approach produces results that rival or surpass the state-of-the-art training-free inference methods.
  arXiv  Detail & Related papers  (2024-10-24T14:52:38Z)
• Decoding-time Realignment of Language Models [44.54462397717971]
  We propose a simple method to explore and evaluate different regularization strengths in aligned models without retraining. DeRa enables control over the degree of alignment, allowing users to smoothly transition between unaligned and aligned models.
  arXiv  Detail & Related papers  (2024-02-05T13:31:28Z)
• One More Step: A Versatile Plug-and-Play Module for Rectifying Diffusion Schedule Flaws and Enhancing Low-Frequency Controls [77.42510898755037]
  One More Step (OMS) is a compact network that incorporates an additional simple yet effective step during inference. OMS elevates image fidelity and harmonizes the dichotomy between training and inference, while preserving original model parameters. Once trained, various pre-trained diffusion models with the same latent domain can share the same OMS module.
  arXiv  Detail & Related papers  (2023-11-27T12:02:42Z)
• DR-Tune: Improving Fine-tuning of Pretrained Visual Models by Distribution Regularization with Semantic Calibration [38.4461170690033]
  We propose a novel fine-tuning framework, namely distribution regularization with semantic calibration (DR-Tune) DR-Tune employs distribution regularization by enforcing the downstream task head to decrease its classification error on the pretrained feature distribution. To alleviate the interference by semantic drift, we develop the semantic calibration (SC) module.
  arXiv  Detail & Related papers  (2023-08-23T10:59:20Z)
• Debiased Fine-Tuning for Vision-language Models by Prompt Regularization [56.48290708901531]
  We present a new paradigm for fine-tuning large-scale vision pre-trained models on downstream task, dubbed Prompt Regularization (ProReg)<n>ProReg uses the prediction by prompting the pretrained model to regularize the fine-tuning.<n>We show the consistently strong performance of ProReg compared with conventional fine-tuning, zero-shot prompt, prompt tuning, and other state-of-the-art methods.
  arXiv  Detail & Related papers  (2023-01-29T11:53:55Z)

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.