# Multi-Step Visual Reasoning with Visual Tokens Scaling and Verification

arXiv: https://arxiv.org/abs/2506.07235 · HTML: https://mlsanity.com/paper/2506.07235/

## Key facts

- Cited by 21 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 18 of its 33 references that are in the map.
- First posted on arXiv 8 Jun 2025.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)
- LLM alignment / RLHF / preference optimisation / [Diffusion and multimodal model alignment](https://mlsanity.com/alignment/diffusion-and-multimodal-model-alignment/)

## Builds on (most cited first, 18 of 18)

- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts](https://mlsanity.com/paper/2310.02255/) (2023)
- [Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations](https://mlsanity.com/paper/2312.08935/) (2023)
- [Are We on the Right Way for Evaluating Large Vision-Language Models?](https://mlsanity.com/paper/2403.20330/) (2024)
- [Shikra: Unleashing Multimodal LLM's Referential Dialogue Magic](https://mlsanity.com/paper/2306.15195/) (2023)
- [MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action](https://mlsanity.com/paper/2303.11381/) (2023)
- [Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V](https://mlsanity.com/paper/2310.11441/) (2023)
- [Visual CoT: Advancing Multi-Modal Language Models with a Comprehensive Dataset and Benchmark for Chain-of-Thought Reasoning](https://mlsanity.com/paper/2403.16999/) (2024)
- [V-STaR: Training Verifiers for Self-Taught Reasoners](https://mlsanity.com/paper/2402.06457/) (2024)
- [MAmmoTH-VL: Eliciting Multimodal Reasoning with Instruction Tuning at Scale](https://mlsanity.com/paper/2412.05237/) (2024)
- [Let's reward step by step: Step-Level reward model as the Navigators for Reasoning](https://mlsanity.com/paper/2310.10080/) (2023)
- [Building Math Agents with Multi-Turn Iterative Preference Learning](https://mlsanity.com/paper/2409.02392/) (2024)
- [Scaffolding Coordinates to Promote Vision-Language Coordination in Large Multi-Modal Models](https://mlsanity.com/paper/2402.12058/) (2024)
- [Improving Large Language Model Fine-tuning for Solving Math Problems](https://mlsanity.com/paper/2310.10047/) (2023)
- [MMFactory: A Universal Solution Search Engine for Vision-Language Tasks](https://mlsanity.com/paper/2412.18072/) (2024)

## Built on by (most cited first, 21 of 21)

- [Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontiers](https://mlsanity.com/paper/2506.23918/) (2025)
- [VABench: A Comprehensive Benchmark for Audio-Video Generation](https://mlsanity.com/paper/2512.09299/) (2025)
- [Beyond Seeing: Evaluating Multimodal LLMs on Tool-Enabled Image Perception, Transformation, and Reasoning](https://mlsanity.com/paper/2510.12712/) (2025)
- [From Perception to Cognition: A Survey of Vision-Language Interactive Reasoning in Multimodal Large Language Models](https://mlsanity.com/paper/2509.25373/) (2025)
- [Simple o3: Towards Interleaved Vision-Language Reasoning](https://mlsanity.com/paper/2508.12109/) (2025)
- [Octopus: Agentic Multimodal Reasoning with Six-Capability Orchestration](https://mlsanity.com/paper/2511.15351/) (2025)
- [Deep But Reliable: Advancing Multi-turn Reasoning for Thinking with Images](https://mlsanity.com/paper/2512.17306/) (2025)
- [Toward Cognitive Supersensing in Multimodal Large Language Model](https://mlsanity.com/paper/2602.01541/) (2026)
- [Jarvis: Towards Personalized AI Assistant via Personal KV-Cache Retrieval](https://mlsanity.com/paper/2510.22765/) (2025)
- [Pix2Fact: When Vision Is Not Enough -- Benchmarking Fine-Grained VQA with Web Verification on High-Resolution Real-World Scenes](https://mlsanity.com/paper/2602.00593/) (2026)
- [Research on World Models Is Not Merely Injecting World Knowledge into Specific Tasks](https://mlsanity.com/paper/2602.01630/) (2026)
- [OpenWorldLib: A Unified Codebase and Definition of Advanced World Models](https://mlsanity.com/paper/2604.04707/) (2026)
- [ATLAS: Agentic or Latent Visual Reasoning? One Word is Enough for Both](https://mlsanity.com/paper/2605.15198/) (2026)
- [MorphoBench: A Benchmark with Difficulty Adaptive to Model Reasoning](https://mlsanity.com/paper/2510.14265/) (2025)
- [Test-time Scaling over Perception: Resolving the Grounding Paradox in Thinking with Images](https://mlsanity.com/paper/2604.11025/) (2026)
- [Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensembles](https://mlsanity.com/paper/2605.22177/) (2026)
- [Visual Prompt Discovery via Semantic Exploration](https://mlsanity.com/paper/2603.16250/) (2026)
- [ROVER: Routing Object-Centric Visual Evidence for Grounded Multi-Image Reasoning](https://mlsanity.com/paper/2605.27959/) (2026)
- [TAEC: Trajectory-Aware Evidence Coordination for Multi-Step Visual RAG](https://mlsanity.com/paper/2609.37349/) (2026)
- [When to Rethink: Learning Multi-Perspective Self-Verification for Vision-Language Models](https://mlsanity.com/paper/2610.07018/) (2026)
- [Thinking with Images without Coordinates](https://mlsanity.com/paper/2610.14265/) (2026)

## Abstract (arXiv)

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.