Conference paper · 2025 · Open access

Multi-step Visual Reasoning with Visual Tokens Scaling and Verification

Tianyi Bai, Zengjie Hu, Fupeng Sun, Jiantao Qiu, Yizhen Jiang, G. He, Bohan Zeng, Conghui He, Binhang Yuan, Wentao Zhang

neural information processing systems, 83024–83062 · Published 1 January 2025 · DOI 10.52202/085713-2504

Open PDFView at the publisherCite this paper

In brief

“Multi-step Visual Reasoning with Visual Tokens Scaling and Verification” is a 2025 conference paper by Tianyi Bai and 9 co-authors, published in neural information processing systems. OpenAlex counts 29 citations as of 19 November 2026, which puts it in the top 1% of works published in 2025 by citations. It is open access: the full text is free to read at arXiv (Cornell University).

Citations
29
Most cited in
2026 (18)
Field-weighted impact
5.82
References
30
Authors
10

Abstract

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.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Bai, T., Hu, Z., Sun, F., Qiu, J., Jiang, Y., He, G., Zeng, B., He, C., Yuan, B., & Zhang, W. (2025). Multi-step Visual Reasoning with Visual Tokens Scaling and Verification. neural information processing systems, 83024–83062. https://doi.org/10.52202/085713-2504

MLA 9

Bai, Tianyi, et al. “Multi-step Visual Reasoning with Visual Tokens Scaling and Verification.” neural information processing systems, 2025, pp. 83024–83062. https://doi.org/10.52202/085713-2504.

Chicago

Bai, Tianyi, Zengjie Hu, Fupeng Sun, Jiantao Qiu, Yizhen Jiang, G. He, Bohan Zeng, Conghui He, Binhang Yuan, and Wentao Zhang. 2025. “Multi-step Visual Reasoning with Visual Tokens Scaling and Verification.” neural information processing systems: 83024–83062. https://doi.org/10.52202/085713-2504.

Harvard

Bai, T. et al. (2025) ‘Multi-step Visual Reasoning with Visual Tokens Scaling and Verification’, neural information processing systems, pp. 83024–83062. doi:10.52202/085713-2504.

IEEE

T. Bai et al., “Multi-step Visual Reasoning with Visual Tokens Scaling and Verification,” neural information processing systems, pp. 83024–83062, 2025, doi: 10.52202/085713-2504.

BibTeX

@inproceedings{bai2025multi,
  title     = {{Multi-step Visual Reasoning with Visual Tokens Scaling and Verification}},
  author    = {Bai, Tianyi and Hu, Zengjie and Sun, Fupeng and Qiu, Jiantao and Jiang, Yizhen and He, G. and Zeng, Bohan and He, Conghui and Yuan, Binhang and Zhang, Wentao},
  booktitle = {neural information processing systems},
  year      = {2025},
  pages     = {83024--83062},
  doi       = {10.52202/085713-2504},
  url       = {https://doi.org/10.52202/085713-2504}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications

  • multimodal large language models
  • iterative inference
  • Markov decision process
  • direct preference optimization
  • visual question answering

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Bai, T., Hu, Z., Sun, F., Qiu, J., Jiang, Y., He, G., Zeng, B., He, C., Yuan, B., & Zhang, W. (2025). Multi-step Visual Reasoning with Visual Tokens Scaling and Verification. neural information processing systems, 83024–83062. https://doi.org/10.52202/085713-2504

How many times has this paper been cited?

29 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at arXiv (Cornell University): sign in and open the PDF from this page.

Who wrote this paper?

Tianyi Bai, Zengjie Hu, Fupeng Sun, Jiantao Qiu, Yizhen Jiang, G. He, Bohan Zeng, Conghui He, Binhang Yuan, Wentao Zhang.

What is the DOI of this paper?

10.52202/085713-2504, permanently at https://doi.org/10.52202/085713-2504. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.