paper-with-me

Papers

Diffusion Models, Denoiser Architecture and Creativity

2026-05-13 · Itamar Levine, Yair Weiss arxiv

The creativity of diffusion models refers to their ability to generate highly realistic images that are different from their training data. Creativity is somewhat surprising since it is known that if the denoiser used in the diffusion model is the Bayes optimal denoiser for a given training set, then the model will simply copy the training samples. In this paper we present empirical and theoretical results that suggest that creativity in diffusion models is due to an interaction between the denoiser architecture and the target distribution. Theoretically, we give explicit forms for the distribution of generated samples as a function of the target distribution and the denoiser architecture for three different denoiser architectures (linear, polynomial, bottleneck). Empirically, we show that small changes in the popular UNET denoiser architecture leads to very different forms of creativity, and these small changes often yield samples that are highly nonrealistic. Taken together, our results show that diffusion models will only be successful if the inductive bias of the denoiser architecture is in strong alignment with the true target distribution.

📄 PDF Abstract BibTeX arXiv:2605.16415

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Towards a Mechanistic Explanation of Diffusion Model Generalization

2024-11-28 · Matthew Niedoba, Berend Zwartsenberg, Kevin Murphy, Frank Wood

We propose a mechanism for diffusion generalization based on local denoising operations. Through analysis of network and empirical denoisers, we identify local inductive biases in diffusion models. We demonstrate that lo…

Denoising

Softly Constrained Denoisers for Diffusion Models Applied to Partial Differential Equations

2025-12-17 · Victor M. Yeom-Song, Severi Rissanen, Arno Solin, Samuel Kaski 외 arxiv

Diffusion models have become a powerful generative prior for solutions of partial differential equations (PDEs). Existing approaches enforce physical constraints either by adding the PDE residuals as loss regularizers or…

Muninn: Your Trajectory Diffusion Model But Faster

2026-05-11 · Gokul Puthumanaillam, Hao Jiang, Ruben Hernandez, Jose Fuentes 외 arxiv

Diffusion-based trajectory planners can synthesize rich, multimodal robot motions, but their iterative denoising makes online planning and control prohibitively slow. Existing accelerations either modify the sampler or c…

Qua$^2$SeDiMo: Quantifiable Quantization Sensitivity of Diffusion Models

2024-12-19 · Keith G. Mills, Mohammad Salameh, Ruichen Chen, Negar Hassanpour 외

Diffusion Models (DM) have democratized AI image generation through an iterative denoising process. Quantization is a major technique to alleviate the inference cost and reduce the size of DM denoiser networks. However, …

DenoisingImage GenerationQuantizationSensitivity

Free Hunch: Denoiser Covariance Estimation for Diffusion Models Without Extra Costs

2024-10-15 · Severi Rissanen, Markus Heinonen, Arno Solin

The covariance for clean data given a noisy observation is an important quantity in many conditional generation methods for diffusion models. Current methods require heavy test-time computation, altering the standard dif…