Multi-Step Visual Reasoning with Visual Tokens Scaling and Verification
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915333163450368 |
|---|---|
| author | Bai, Tianyi Hu, Zengjie Sun, Fupeng Qiu, Jiantao Jiang, Yizhen He, Guangxin Zeng, Bohan He, Conghui Yuan, Binhang Zhang, Wentao |
| author_facet | Bai, Tianyi Hu, Zengjie Sun, Fupeng Qiu, Jiantao Jiang, Yizhen He, Guangxin Zeng, Bohan He, Conghui Yuan, Binhang Zhang, Wentao |
| contents | Multi-modal large language models (MLLMs) have achieved remarkable capabilities by integrating visual perception with language understanding, enabling applications such as image-grounded dialogue, visual question answering, and scientific analysis. However, most MLLMs adopt a static inference paradigm, encoding the entire image into fixed visual tokens upfront, which limits their ability to iteratively refine understanding or adapt to context during inference. This contrasts sharply with human perception, which is dynamic, selective, and feedback-driven. In this work, we introduce a novel framework for inference-time visual token scaling that enables MLLMs to perform iterative, verifier-guided reasoning over visual content. We formulate the problem as a Markov Decision Process, involving a reasoner that proposes visual actions and a verifier, which is trained via multi-step Direct Preference Optimization (DPO), that evaluates these actions and determines when reasoning should terminate. To support this, we present a new dataset, VTS, comprising supervised reasoning trajectories (VTS-SFT) and preference-labeled reasoning comparisons (VTS-DPO). Our method significantly outperforms existing approaches across diverse visual reasoning benchmarks, offering not only improved accuracy but also more interpretable and grounded reasoning processes. These results demonstrate the promise of dynamic inference mechanisms for enabling fine-grained, context-aware visual reasoning in next-generation MLLMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_07235 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Multi-Step Visual Reasoning with Visual Tokens Scaling and Verification Bai, Tianyi Hu, Zengjie Sun, Fupeng Qiu, Jiantao Jiang, Yizhen He, Guangxin Zeng, Bohan He, Conghui Yuan, Binhang Zhang, Wentao Computer Vision and Pattern Recognition Computation and Language Multi-modal large language models (MLLMs) have achieved remarkable capabilities by integrating visual perception with language understanding, enabling applications such as image-grounded dialogue, visual question answering, and scientific analysis. However, most MLLMs adopt a static inference paradigm, encoding the entire image into fixed visual tokens upfront, which limits their ability to iteratively refine understanding or adapt to context during inference. This contrasts sharply with human perception, which is dynamic, selective, and feedback-driven. In this work, we introduce a novel framework for inference-time visual token scaling that enables MLLMs to perform iterative, verifier-guided reasoning over visual content. We formulate the problem as a Markov Decision Process, involving a reasoner that proposes visual actions and a verifier, which is trained via multi-step Direct Preference Optimization (DPO), that evaluates these actions and determines when reasoning should terminate. To support this, we present a new dataset, VTS, comprising supervised reasoning trajectories (VTS-SFT) and preference-labeled reasoning comparisons (VTS-DPO). Our method significantly outperforms existing approaches across diverse visual reasoning benchmarks, offering not only improved accuracy but also more interpretable and grounded reasoning processes. These results demonstrate the promise of dynamic inference mechanisms for enabling fine-grained, context-aware visual reasoning in next-generation MLLMs. |
| title | Multi-Step Visual Reasoning with Visual Tokens Scaling and Verification |
| topic | Computer Vision and Pattern Recognition Computation and Language |
| url | https://arxiv.org/abs/2506.07235 |