V-Thinker: Interactive Thinking with Images

2025-11-06

AI Summary

Current large multimodal models (LMMs) suffer from limited visual tool spaces and task-specific workflows, hindering fine-grained image interaction and long-horizon reasoning. To address this, we propose a vision-centric interactive reasoning paradigm, introducing a “data evolution flywheel” that automatically generates high-quality, multi-difficulty interactive reasoning data. We further design a vision-progressive training curriculum integrating point-level supervision with a two-stage reinforcement learning framework, enabling end-to-end optimization of perceptual alignment and interactive reasoning. Our method substantially improves both general and interactive reasoning capabilities, achieving comprehensive superiority over state-of-the-art LMMs on our expert-validated benchmark VTBench—a newly constructed evaluation suite for visual interactive reasoning. This work establishes a new benchmark and a scalable technical pathway for image-driven interactive cognitive modeling.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Empowering Large Multimodal Models (LMMs) to deeply integrate image interaction with long-horizon reasoning capabilities remains a long-standing challenge in this field. Recent advances in vision-centric reasoning explore a promising"Thinking with Images"paradigm for LMMs, marking a shift from image-assisted reasoning to image-interactive thinking. While this milestone enables models to focus on fine-grained image regions, progress remains constrained by limited visual tool spaces and task-specific workflow designs. To bridge this gap, we present V-Thinker, a general-purpose multimodal reasoning assistant that enables interactive, vision-centric thinking through end-to-end reinforcement learning. V-Thinker comprises two key components: (1) a Data Evolution Flywheel that automatically synthesizes, evolves, and verifies interactive reasoning datasets across three dimensions-diversity, quality, and difficulty; and (2) a Visual Progressive Training Curriculum that first aligns perception via point-level supervision, then integrates interactive reasoning through a two-stage reinforcement learning framework. Furthermore, we introduce VTBench, an expert-verified benchmark targeting vision-centric interactive reasoning tasks. Extensive experiments demonstrate that V-Thinker consistently outperforms strong LMM-based baselines in both general and interactive reasoning scenarios, providing valuable insights for advancing image-interactive reasoning applications.

Problem

• Integrating image interaction with long-horizon reasoning in multimodal models
• Overcoming limited visual tool spaces and task-specific workflow designs
• Advancing image-interactive reasoning through reinforcement learning frameworks

Innovation

• End-to-end reinforcement learning for vision-centric thinking
• Data Evolution Flywheel synthesizes interactive reasoning datasets
• Visual Progressive Training Curriculum with two-stage reinforcement learning