Enregistré dans:
Détails bibliographiques
Auteurs principaux: Xiao, Yicheng, Song, Lin, Chen, Yukang, Luo, Yingmin, Chen, Yuxin, Gan, Yukang, Huang, Wei, Li, Xiu, Qi, Xiaojuan, Shan, Ying
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2505.13031
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913888952385536
author Xiao, Yicheng
Song, Lin
Chen, Yukang
Luo, Yingmin
Chen, Yuxin
Gan, Yukang
Huang, Wei
Li, Xiu
Qi, Xiaojuan
Shan, Ying
author_facet Xiao, Yicheng
Song, Lin
Chen, Yukang
Luo, Yingmin
Chen, Yuxin
Gan, Yukang
Huang, Wei
Li, Xiu
Qi, Xiaojuan
Shan, Ying
contents Recent text-to-image systems face limitations in handling multimodal inputs and complex reasoning tasks. We introduce MindOmni, a unified multimodal large language model that addresses these challenges by incorporating reasoning generation through reinforcement learning. MindOmni leverages a three-phase training strategy: i) design of a unified vision language model with a decoder-only diffusion module, ii) supervised fine-tuning with Chain-of-Thought (CoT) instruction data, and iii) our proposed Reasoning Generation Policy Optimization (RGPO) algorithm, utilizing multimodal feedback to effectively guide policy updates. Experimental results demonstrate that MindOmni outperforms existing models, achieving impressive performance on both understanding and generation benchmarks, meanwhile showcasing advanced fine-grained reasoning generation capabilities, especially with mathematical reasoning instruction. All codes will be made public at https://github.com/TencentARC/MindOmni
format Preprint
id arxiv_https___arxiv_org_abs_2505_13031
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MindOmni: Unleashing Reasoning Generation in Vision Language Models with RGPO
Xiao, Yicheng
Song, Lin
Chen, Yukang
Luo, Yingmin
Chen, Yuxin
Gan, Yukang
Huang, Wei
Li, Xiu
Qi, Xiaojuan
Shan, Ying
Artificial Intelligence
Recent text-to-image systems face limitations in handling multimodal inputs and complex reasoning tasks. We introduce MindOmni, a unified multimodal large language model that addresses these challenges by incorporating reasoning generation through reinforcement learning. MindOmni leverages a three-phase training strategy: i) design of a unified vision language model with a decoder-only diffusion module, ii) supervised fine-tuning with Chain-of-Thought (CoT) instruction data, and iii) our proposed Reasoning Generation Policy Optimization (RGPO) algorithm, utilizing multimodal feedback to effectively guide policy updates. Experimental results demonstrate that MindOmni outperforms existing models, achieving impressive performance on both understanding and generation benchmarks, meanwhile showcasing advanced fine-grained reasoning generation capabilities, especially with mathematical reasoning instruction. All codes will be made public at https://github.com/TencentARC/MindOmni
title MindOmni: Unleashing Reasoning Generation in Vision Language Models with RGPO
topic Artificial Intelligence
url https://arxiv.org/abs/2505.13031