Metraipolis › archive › dc_afa2d8cf8bbb

Latent Reasoning in Continuous Space for Unified Multimodal Models — arXiv:2610.42149

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.42149
License (as archived): CC0-1.0
Archived: 2026-11-03T07:55:29.397Z
SHA-256 of the archived text: a66254fbb1680981ef6ed86e795d5a11598e991f6ccc335f158a3f713d2880e6
Stable URL of this page: https://metraipolis.com/documento/dc_afa2d8cf8bbb

How to cite

Latent Reasoning in Continuous Space for Unified Multimodal Models — arXiv:2610.42149 — arXiv (en), https://arxiv.org/abs/2610.42149 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_afa2d8cf8bbb (Metraipolis archive dc_afa2d8cf8bbb, sha256 a66254fbb1680981ef6ed86e795d5a11598e991f6ccc335f158a3f713d2880e6)

BibTeX

@misc{dcafa2d8cf8bbb,
  title = {Latent Reasoning in Continuous Space for Unified Multimodal Models — arXiv:2610.42149},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.42149},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_afa2d8cf8bbb; sha256 a66254fbb1680981ef6ed86e795d5a11598e991f6ccc335f158a3f713d2880e6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.42149. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Unified Multimodal Models (UMMs) achieve remarkable success on diverse multimodal understanding and generation tasks, yet they still struggle to effectively leverage language-based chain-of-thought reasoning for image generation. As a result, existing reasoning-augmented UMMs often fail to inherit the test-time scaling and fine-grained controllability observed in large language models. We attribute this failure to a fundamental modality mismatch: reasoning is performed in a discrete token space, whereas image generation operates in a continuous space. To bridge this gap, we introduce a Latent Reasoning framework in Continuous space (LARC), which alleviates the token-level bottleneck by reasoning over continuous hidden states. Specifically, the training framework consists of two stages: (i) curriculum supervised fine-tuning (SFT), and (ii) information-gain reinforcement learning (RL). First, the curriculum SFT stage gradually converts token-level reasoning into latent reasoning. Subsequently, the information-gain RL stage self-evolves latent reasoning without ground truth supervision. Experiments across diverse text-to-image generation benchmarks show that LARC consistently improves generation quality, prompt following, and compositional fidelity. Notably, LARC achieves state-of-the-art performance on GenEval with a 7%p gain over BAGEL and demonstrates clear test-time scaling behavior, outperforming token-based generation under both parallel and sequential test-time scaling strategies.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Computation and Language (cs.CL); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.42149 — 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 2026-11-03T07:49:46.038Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).