| title | Latent Reasoning in Continuous Space for Unified Multimodal Models |
|---|---|
| authors | Byungwoo Jeon, Yoonwoo Jeong, Hyunseok Lee, Minsu Cho, Jinwoo Shin |
| date | 2026-10-30 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2610.42149 |
| affiliation | KAIST AI |
| upvotes | 25 (Hugging Face Daily Papers, 2026-11-03) |
| links | abs pdf hf code |
| abstract | 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. |
cite
@misc{jeon2026latent,
title={Latent Reasoning in Continuous Space for Unified Multimodal Models},
author={Byungwoo Jeon and Yoonwoo Jeong and Hyunseok Lee and Minsu Cho and Jinwoo Shin},
year={2026},
eprint={2610.42149},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2610.42149}
}