SenseNova-MARS: 通过强化学习赋能多模态代理推理与搜索
SenseNova-MARS: Empowering Multimodal Agentic Reasoning and Search via Reinforcement Learning
AI总结：
SenseNova-MARS通过强化学习赋能多模态代理推理与搜索，提升视觉-语言模型在复杂视觉任务中的表现。
摘要：
尽管视觉-语言模型（VLMs）可以通过代理推理解决复杂任务，但其能力仍主要受限于文本导向的链式推理或孤立工具调用。它们无法表现出人类所需的无缝结合动态工具操作与连续推理的能力，尤其是在需要协调外部工具如搜索和图像裁剪的知识密集型和视觉复杂场景中。在本工作中，我们引入了SenseNova-MARS，一种新颖的多模态代理推理与搜索框架，通过强化学习（RL）使VLMs具备交错的视觉推理和工具使用能力。具体而言，SenseNova-MARS动态整合图像搜索、文本搜索和图像裁剪工具，以应对细粒度和知识密集型的视觉理解挑战。在RL阶段，我们提出了一种批归一化组序列策略优化（BN-GSPO）算法，以提高训练稳定性并提升模型调用工具和有效推理的能力。为了全面评估代理VLMs在复杂视觉任务上的能力，我们引入了HR-MMSearch基准测试，这是首个以搜索为导向的基准测试，由高分辨率图像和知识密集型、搜索驱动的问题组成。实验表明，SenseNova-MARS在开源搜索和细粒度图像理解基准测试中实现了最先进的性能。具体而言，在以搜索为导向的基准测试中，SenseNova-MARS-32B在MMSearch上得分为74.3，在HR-MMSearch上得分为54.4，超过了Gemini-3-Pro和GPT-5.2等专有模型。SenseNova-MARS通过提供有效且稳健的工具使用能力，代表了代理VLMs的一个有前景的进展。为了促进该领域进一步的研究，我们将发布所有代码、模型和数据集。
英文摘要：
While Vision-Language Models (VLMs) can solve complex tasks through agentic reasoning, their capabilities remain largely constrained to text-oriented chain-of-thought or isolated tool invocation. They fail to exhibit the human-like proficiency required to seamlessly interleave dynamic tool manipulation with continuous reasoning, particularly in knowledge-intensive and visually complex scenarios that demand coordinated external tools such as search and image cropping. In this work, we introduce SenseNova-MARS, a novel Multimodal Agentic Reasoning and Search framework that empowers VLMs with interleaved visual reasoning and tool-use capabilities via reinforcement learning (RL). Specifically, SenseNova-MARS dynamically integrates the image search, text search, and image crop tools to tackle fine-grained and knowledge-intensive visual understanding challenges. In the RL stage, we propose the Batch-Normalized Group Sequence Policy Optimization (BN-GSPO) algorithm to improve the training stability and advance the model's ability to invoke tools and reason effectively. To comprehensively evaluate the agentic VLMs on complex visual tasks, we introduce the HR-MMSearch benchmark, the first search-oriented benchmark composed of high-resolution images with knowledge-intensive and search-driven questions. Experiments demonstrate that SenseNova-MARS achieves state-of-the-art performance on open-source search and fine-grained image understanding benchmarks. Specifically, on search-oriented benchmarks, SenseNova-MARS-32B scores 74.3 on MMSearch and 54.4 on HR-MMSearch, surpassing proprietary models such as Gemini-3-Pro and GPT-5.2. SenseNova-MARS represents a promising step toward agentic VLMs by providing effective and robust tool-use capabilities. To facilitate further research in this field, we will release all code, models, and datasets.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
