| title | SEGA: Spectral-Energy Guided Attention for Resolution Extrapolation in Diffusion Transformers |
|---|---|
| authors | Javad Rajabi, Kimia Shaban, Koorosh Roohi, David B. Lindell, Babak Taati |
| date | 2026-05-21 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2605.22668 |
| affiliation | University of Toronto Computer Science |
| upvotes | 43 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Diffusion transformers (DiTs) have emerged as a dominant architecture for text-to-image generation, yet their performance drops when generating at resolutions beyond their training range. Existing training-free approaches mitigate this by modifying inference-time attention behavior, often through Rotary Position Embeddings (RoPE) extrapolation combined with attention scaling. However, these strategies apply a uniform and content-agnostic scaling across RoPE components with distinct frequency characteristics, inducing a trade-off between preserving global structure and recovering fine detail. We introduce SEGA, a training-free method that dynamically scales attention across RoPE components according to the latent's spatial-frequency structure at each denoising step. This adaptive scaling improves both structural coherence and fine-detail fidelity. Experiments show that SEGA consistently improves high-resolution synthesis across multiple target resolutions, outperforming state-of-the-art training-free baselines. |
cite
@misc{rajabi2026sega,
title={SEGA: Spectral-Energy Guided Attention for Resolution Extrapolation in Diffusion Transformers},
author={Javad Rajabi and Kimia Shaban and Koorosh Roohi and David B. Lindell and Babak Taati},
year={2026},
eprint={2605.22668},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2605.22668}
}