Metraipolis › archive › dc_2c14c70fe3fe

MoLE: Mixture of Latent Experts for Complementary Visual Reasoning — arXiv:2610.01917

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.01917
License (as archived)
CC0-1.0
Archived
2026-10-06T12:15:31.472Z
SHA-256 of the archived text
c445231d210448535bb67d90f32fe017fdf5e6283be29fc7496276330febdff7
Stable URL of this page
https://metraipolis.com/documento/dc_2c14c70fe3fe

How to cite

MoLE: Mixture of Latent Experts for Complementary Visual Reasoning — arXiv:2610.01917 — arXiv (en), https://arxiv.org/abs/2610.01917 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2c14c70fe3fe (Metraipolis archive dc_2c14c70fe3fe, sha256 c445231d210448535bb67d90f32fe017fdf5e6283be29fc7496276330febdff7)
BibTeX
@misc{dc2c14c70fe3fe,
  title = {MoLE: Mixture of Latent Experts for Complementary Visual Reasoning — arXiv:2610.01917},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.01917},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2c14c70fe3fe; sha256 c445231d210448535bb67d90f32fe017fdf5e6283be29fc7496276330febdff7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.01917. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Latent visual reasoning equips vision--language models with continuous intermediate states that can process visual evidence without explicit textual reasoning traces or repeated image operations. However, existing methods often allow multiple latent tokens to access the same visual evidence through shared value projections, providing no mechanism for them to extract complementary visual information; simply increasing the latent budget can therefore yield redundant latent representations. We argue that effective latent reasoning should encourage different latent tokens to extract complementary visual information, and thereby act as specialized visual experts. Based on this insight, we propose MoLE, a Mixture of Latent Experts framework that controls both what visual evidence each latent visual expert observes and how it transforms that evidence. MoLE isolates latent visual experts during evidence extraction and uses dedicated latent summary experts to aggregate the complementary representations of latent visual experts. A two-stage training pipeline first forces visual evidence through this latent pathway and then restores direct visual access, requiring neither predefined expert roles nor intermediate visual targets. Across five visual reasoning benchmarks, MoLE achieves an average score of 78.6, outperforming data-matched supervised fine-tuning by 4.9 and the strongest evaluated latent visual reasoning baseline at the same latent budget by 3.6. Representation analyses show lower latent-state similarity and more diverse visual attention, while masking the latent pathway reduces average performance by 9.2. These results demonstrate that specializing latent computation is more effective than merely increasing the number of latent tokens.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-02 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.01917 — 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 2026-10-06T10:21:28.228Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).