Metraipolis › archive › dc_d41b76a38c70

Pooling Representation Autoencoders for Efficient Diffusion — arXiv:2610.09242

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.09242
License (as archived)
CC0-1.0
Archived
2026-10-08T12:47:02.390Z
SHA-256 of the archived text
bd0d4d3c3235a00b1092d2bb0acec4bc3714e270f6495a1c85d77727531718a0
Stable URL of this page
https://metraipolis.com/documento/dc_d41b76a38c70

How to cite

Pooling Representation Autoencoders for Efficient Diffusion — arXiv:2610.09242 — arXiv (en), https://arxiv.org/abs/2610.09242 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d41b76a38c70 (Metraipolis archive dc_d41b76a38c70, sha256 bd0d4d3c3235a00b1092d2bb0acec4bc3714e270f6495a1c85d77727531718a0)
BibTeX
@misc{dcd41b76a38c70,
  title = {Pooling Representation Autoencoders for Efficient Diffusion — arXiv:2610.09242},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09242},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d41b76a38c70; sha256 bd0d4d3c3235a00b1092d2bb0acec4bc3714e270f6495a1c85d77727531718a0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09242. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Representation Autoencoders (RAEs) generate images from pre-trained visual fea- tures, but their dense token grids make generative modeling expensive. Motivated by local feature correlations, we introduce PoolDINO, a learned affine pooling operator that merges neighboring tokens. Training the pooling operator jointly with the RGB decoder preserves the standard two-stage RAE procedure without a separate feature auto-encoder. On ImageNet-256, 4x token compression retains comparable generation quality under internal guidance, while 16x compression trades some quality for greater efficiency. At a fixed budget of 100 sampling steps, latent-sampling throughput increases by 3.7x and 9.0x, respectively, relative to the unpooled baseline. Classification and dense prediction evaluations show that comparable guided generation quality can coexist with weaker performance on other tasks.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.09242 — 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 2026-10-08T12:18:15.321Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).