Metraipolis › archive › dc_db0f11485eba

RefAdapt-DiT: Adaptive Joint Attention for Reference-Conditioned Diffusion Transformers — arXiv:2609.32415

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.32415
License (as archived)
CC0-1.0
Archived
2026-10-07T23:50:54.311Z
SHA-256 of the archived text
4ec9621ed0138ef5241ea6193c34bc4af4d66e376940e1e47302ecd05945c73f
Stable URL of this page
https://metraipolis.com/documento/dc_db0f11485eba

How to cite

RefAdapt-DiT: Adaptive Joint Attention for Reference-Conditioned Diffusion Transformers — arXiv:2609.32415 — arXiv (en), https://arxiv.org/abs/2609.32415 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_db0f11485eba (Metraipolis archive dc_db0f11485eba, sha256 4ec9621ed0138ef5241ea6193c34bc4af4d66e376940e1e47302ecd05945c73f)
BibTeX
@misc{dcdb0f11485eba,
  title = {RefAdapt-DiT: Adaptive Joint Attention for Reference-Conditioned Diffusion Transformers — arXiv:2609.32415},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.32415},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_db0f11485eba; sha256 4ec9621ed0138ef5241ea6193c34bc4af4d66e376940e1e47302ecd05945c73f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32415. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Diffusion Transformers (DiTs) have become the standard backbone for high-quality generative modeling, yet deploying them in conditional generation tasks remains computationally prohibitive because bidirectional joint attention repeatedly processes large reference streams. While existing optimization schemes mitigate generic temporal redundancy, they typically rely on coarse-grained static reuse and overlook the distinct dynamics of references and targets. Specifically, we observe that reference representations often evolve slowly along the generation trajectory, while the target often assigns little attention mass to them; reference drift and this target-to-reference exposure jointly shape how strongly stale reference states affect the target. To exploit these patterns, we introduce \RefAdapt, a training-free framework for adaptive control of joint attention between references and targets. Instead of rigid static strategies, \RefAdapt combines consecutive target-Q change with previously observed target-to-reference attention mass to control reference computation adaptively at block granularity. Under ultra-few-step settings, \RefAdapt enables speedups of up to $2.097\times$ on 4-step MiniMax H3 and $3.54\times$ on 8-step Qwen Image Edit, while maintaining comparable visual quality.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.32415 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T23:29:27.038Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).