LogT:用逻辑链驱动的图像思考提升视觉搜索能力
LogT: Logically Think with Images for Visual Search
这篇论文有意思的地方是反过来做:不堆 RL 算力,而是把 SFT 数据的逻辑质量做扎实。Qwen2.5-VL-7B 微调后只用原来 1/120 的 GPU 时间就超过了此前 SFT+RL 的最强结果。做视觉 Agent 或多模态训练的团队值得看他们的数据构造方法。
论文提出 LogT(Logically Think with Images),一个全自动 SFT 数据引擎,通过逻辑链引导的推理生成高质量的视觉搜索训练数据,包含有挑战性的视觉接地问题和结构化的工具调用轨迹。作者将研究重点从提升 RL 复杂度转向 SFT 数据质量,认为后者才是关键瓶颈。
实验中,研究团队用 LogT 数据集微调 Qwen2.5-VL-7B 和 InternVL3.5-8B 两个基座模型,并在 7 个基准测试集上评估。基于 Qwen2.5-VL-7B 的 LogT-SFT 仅用 3.1 倍更少的数据和超过 120 倍更少的 GPU 小时,就超过了此前用 SFT 和 RL 训练的最优方法。在 LogT-SFT 之上再加一个无需工具特定设计的简单 RL 流程,模型比此前复杂 RL 方案高出 9.59 分。
除性能提升外,该模型的工具调用准确率和效率更高,推理轨迹更结构化、逻辑更一致。这对资源有限的开发者有直接参考价值:与其堆 RL 训练算力,不如先把 SFT 数据的逻辑质量做好。
LogT: Logically Think with Images for Visual Search
Thinking with images empowers vision-language models (VLMs) to use an external Zoom-in tool for fine-grained visual search. Recent works employ supervised fine-tuning (SFT) and reinforcement learning (RL) with tool-specific designs to enable such capabilities. However, achieving logical reasoning alongside accurate and efficient tool use remains a significant challenge. To address this, we shift the focus away from increasing RL complexity and instead identify SFT data quality as a key bottleneck. We then propose LogT (short for Logically Think with Images), a fully automated SFT data engine that explicitly enforces logic-chain-guided reasoning. LogT generates an SFT dataset featuring challenging, visually grounded questions and structured tool-use traces with human-like logic. We validate our approach by fine-tuning two base models, Qwen2.5-VL-7B and InternVL3.5-8B, on the dataset and evaluating them across seven benchmark test sets. We show that LogT-SFT based on Qwen2.5-VL-7B outperforms the state-of-the-art trained with both SFT and RL, despite using 3.1$\times$ less data and $>120\times$ fewer GPU hours. Furthermore, applying a simple RL recipe without tool-specific design on LogT-SFT yields a model that surpasses prior methods with complex RL setups by a substantial 9.59-point margin. In addition to strong performance gains, our model exhibits higher tool-use accuracy and efficiency, as well as more structured and logically consistent reasoning traces, directly validating the effectiveness of our approach.
This report is categorized under AITOP's "Papers" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Qwen2.5-VL, InternVL3.5, visual search, SFT.
Source attribution: curated from arXiv cs.LG (@Yanjun Fu, Quanzeng You, Jiadong Guo, Yujie Lu, Sai Charitha Akula, Haiping Wu, Miao Liu, Sanghamitra Dutta, Lu Yuan).