Abstract
Large Vision-Language Models (LVLMs) have demonstrated remarkable success in a broad range of vision-language tasks, such as general visual question answering and optical character recognition (OCR). However, their performance on perception-centric tasks -- such as object detection, semantic segmentation, and depth estimation -- remains significantly inferior to that of task-specific expert models. For example, Qwen2.5-VL-7B-Instruct achieves only 19% mAP on COCO2017 val, particularly struggling with dense scenes and small object recall. In this work, we introduce Chain-of-Thought for Detection (CoT4Det), a simple but efficient strategy that reformulates perception tasks into three interpretable steps: classification, counting, and grounding -- each more naturally aligned with the reasoning capabilities of LVLMs. Extensive experiments demonstrate that our method significantly improves perception performance without compromising general vision language capabilities. With a standard Qwen2.5-VL-7B-Instruct, CoT4Det boosts mAP from 19.0% to 33.0% on COCO2017 val and achieves competitive results across a variety of perception benchmarks, outperforming baselines by +2% on RefCOCO series and 19% on Flickr30k entities.
Abstract (translated)
大型视觉-语言模型（LVLMs）在广泛的视觉-语言任务中展示了显著的成功，例如通用的视觉问答和光学字符识别（OCR）。然而，在以感知为中心的任务——如物体检测、语义分割和深度估计上，它们的表现仍然明显低于特定任务的专家模型。例如，Qwen2.5-VL-7B-Instruct在COCO2017验证集上的mAP仅为19%，尤其在密集场景和小目标召回方面表现不佳。 在这项工作中，我们介绍了用于检测的Chain-of-Thought方法（CoT4Det），这是一种简单而高效的策略，它将感知任务重新表述为三个可解释的步骤：分类、计数和定位——每个步骤都更自然地与LVLMs的推理能力相匹配。广泛的实验表明，我们的方法在不牺牲通用视觉语言能力的情况下显著提高了感知性能。使用标准的Qwen2.5-VL-7B-Instruct模型，CoT4Det将COCO2017验证集上的mAP从19.0%提升到33.0%，并在各种感知基准测试中取得了具有竞争力的结果，在RefCOCO系列上比基线高出+2%，在Flickr30k实体数据集上高出了19%。
URL
https://arxiv.org/abs/2512.06663
