V-Thinker：基于图像的交互式思考

摘要

赋能大型多模态模型（LMM）实现图像交互与长程推理能力的深度融合，始终是该领域长期存在的挑战。近期以视觉为核心的推理研究探索出极具前景的"图像化思考"范式，标志着从图像辅助推理向图像交互式思维的转变。尽管这一里程碑式进展使模型能聚焦细粒度图像区域，但有限的视觉工具空间与任务定制化流程设计仍制约着进一步发展。为突破此局限，我们提出V-Thinker——一种通过端到端强化学习实现交互式视觉中心化思维的通用多模态推理助手。该框架包含两大核心组件：（1）数据进化飞轮，可沿多样性、质量与难度三维度自动合成、演进并验证交互式推理数据集；（2）视觉渐进式训练课程，先通过点级监督实现感知对齐，再经由两阶段强化学习框架融合交互推理。此外，我们推出经专家验证的VTBench基准测试集，专门针对视觉中心化交互推理任务。大量实验表明，V-Thinker在通用推理与交互推理场景中均持续超越基于LMM的强基线模型，为推进图像交互式推理应用提供了宝贵洞见。

English

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.

V-Thinker：基于图像的交互式思考

V-Thinker: Interactive Thinking with Images

摘要

Support