MindOmni: Unleashing Reasoning Generation in Vision Language Models with RGPO

TL;DR

MindOmni is a unified multimodal large language model that enhances reasoning and generation capabilities in vision-language tasks through a novel training strategy including RGPO reinforcement learning.

Contribution

It introduces a new training approach with RGPO for multimodal models, improving reasoning and generation performance over existing systems.

Findings

Outperforms existing models on understanding benchmarks

Demonstrates advanced reasoning, especially in mathematical tasks

Achieves superior multimodal reasoning and generation capabilities

Abstract

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