Multimodal reasoning improved with unified state based iterative method
From Stages to States: Unified Iterative Reasoning for MLLMs
Artificial Intelligence
Summary
Solving complex problems with AI that can see and read often breaks the process into separate steps, which can cause mistakes and gaps. The authors designed ReState, which treats reasoning as a continuous state that updates itself, allowing the AI to learn from its own thinking and fix errors along the way. They also created a new dataset to help train this approach across different visual problem types. Their experiments show this method works well with various AI models and reasoning challenges.
What this means in practice
- •For multimodal ai developers: Integrate a unified reasoning state model to improve error correction and iterative reasoning in AI systems that process both images and text.
- •For visual reasoning system builders: Use the VersaCoT-10K dataset to train and refine reasoning capabilities across diverse visual domains, boosting system robustness.
- •For automated quality control teams: Improve automated inspection systems by applying unified iterative reasoning to detect and correct interpretation errors without separate correction modules.$Commercial implications: Enables selling enhanced AI-powered visual inspection tools that self-correct reasoning errors, improving reliability in manufacturing.
Authors
Chuang Yu, Xuefei Ma, Yuanhao Feng, Wuxiong Zeng, Bo Wang, Minghong Cai, Dongming Wu, Fei Zhou, Chen Tang, Wenwen Tong, Lewei Lu, Xiangyu Yue
Abstract
Multimodal Chain-of-Thought (MCoT) substantially enhances the complex reasoning capabilities of Multimodal Large Language Models (MLLMs) by explicitly modeling intermediate reasoning processes. However, most learning-based MCoT methods follow a stage-centric paradigm, where different reasoning behaviors are assigned to predefined stages and optimized separately. This results in fragmented reasoning and a training-inference reasoning-state gap. To address these limitations, we propose ReState, a unified iterative reasoning framework that reformulates multimodal reasoning from a stage-driven pipeline into a state-driven dynamic process. It comprises two complementary components: Unified Reasoning State Modeling (URSM) and Closed-Loop Reasoning-State Learning (CRSL). URSM unifies reasoning construction, correct reasoning preservation, and erroneous reasoning correction as reasoning-state transitions within a shared state space, enabling these heterogeneous reasoning behaviors to be learned jointly. CRSL further incorporates self-generated historical reasoning states during training and recursively updates the current state during inference, thereby mitigating the reasoning-state gap and enabling iterative self-refinement without requiring an additional reflection module, critic, or independent correction stage. Furthermore, we construct VersaCoT-10K, a high-quality CoT dataset spanning eight visual reasoning domains, which provides diverse supervision for multi-domain reasoning and reasoning-state refinement. Extensive experiments demonstrate that ReState exhibits strong adaptability across different MLLM backbones and achieves strong performance on multiple public reasoning benchmarks, supporting the shift from stage-centric reasoning toward state-centric reasoning. Our data and code are available at https://github.com/YuChuang1205/ReState