← Search

Itir Onal Ertugrul

2 accepted papers

2025

DCTdiff: Intriguing Properties of Image Generative Modeling in the DCT Space

ICML 2025poster

This paper explores image modeling from the frequency space and introduces DCTdiff, an end-to-end diffusion generative paradigm that efficiently models images in the discrete cosine transform (DCT) space. We investigate the design space of DCTdiff and reveal the key design factors. Experiments on di…

2024

Elucidating the Exposure Bias in Diffusion Models

ICLR 2024poster

Diffusion models have demonstrated impressive generative capabilities, but their exposure bias problem, described as the input mismatch between training and sampling, lacks in-depth exploration. In this paper, we investigate the exposure bias problem in diffusion models by first analytically modelli…