Saved in:
Bibliographic Details
Main Authors: Tokhchukov, Danil, Mirzoeva, Aysel, Kuznetsov, Andrey, Sobolev, Konstantin
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.24800
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • In this paper, we uncover the hidden potential of Diffusion Transformers (DiTs) to significantly enhance generative tasks. Through an in-depth analysis of the denoising process, we demonstrate that introducing a single learned scaling parameter can significantly improve the performance of DiT blocks. Building on this insight, we propose Calibri, a parameter-efficient approach that optimally calibrates DiT components to elevate generative quality. Calibri frames DiT calibration as a black-box reward optimization problem, which is efficiently solved using an evolutionary algorithm and modifies just ~100 parameters. Experimental results reveal that despite its lightweight design, Calibri consistently improves performance across various text-to-image models. Notably, Calibri also reduces the inference steps required for image generation, all while maintaining high-quality outputs.