arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2510.04401cs.CVcs.AI
收藏

你的视觉语言模型甚至数不到20:揭示VLM在组合计数中的失败

Your Vision-Language Model Can't Even Count to 20: Exposing the Failures of VLMs in Compositional Counting

Guilin University of Electronic Technology(桂林电子科技大学);The Ohio State University(俄亥俄州立大学);University of Wisconsin-Madison(威斯康星大学麦迪逊分校);University of California, Berkeley(加州大学伯克利分校)

Xuyang Guo, Zekai Huang, Zhenmei Shi, Zhao Song, Jiahao Zhang

AI总结:

本文针对视觉语言模型(VLM)的物体计数能力问题,提出仅含基础几何形状的VLMCountBench基准,通过受控实验发现VLM单类形状计数可靠但组合计数存在大量失败,揭示了其当前的核心局限。

摘要:

视觉语言模型(VLM)凭借从网络大规模视觉语言数据训练中获得的出色能力,已成为当今AI社区的核心焦点。这类模型在图像理解、视频理解、复杂视觉推理和具身AI等多种任务中展现出强劲性能。尽管取得了这些显著成就,一个基本问题仍然存在:VLM能否正确计数物体?本文提出了一个简单却有效的基准VLMCountBench,该基准在极简设置下设计,仅包含基础几何形状(如三角形、圆形)及其组合,专门聚焦计数任务,不受其他因素干扰。我们采用严格的自变量控制,在受控消融实验中系统研究了颜色、大小和提示优化等简单属性的影响。实证结果表明,当仅存在一种形状类型时,VLM能够可靠计数,但当多种形状类型组合时(即组合计数),它们会出现大量失败。这一发现凸显了当前VLM的一项根本性实证局限,也为未来研究指明了重要方向。

英文摘要:

Vision-Language Models (VLMs) have become a central focus of today's AI community, owing to their impressive abilities gained from training on large-scale vision-language data from the Web. These models have demonstrated strong performance across diverse tasks, including image understanding, video understanding, complex visual reasoning, and embodied AI. Despite these noteworthy successes, a fundamental question remains: Can VLMs count objects correctly? In this paper, we introduce a simple yet effective benchmark, VLMCountBench, designed under a minimalist setting with only basic geometric shapes (e.g., triangles, circles) and their compositions, focusing exclusively on counting tasks without interference from other factors. We adopt strict independent variable control and systematically study the effects of simple properties such as color, size, and prompt refinement in a controlled ablation. Our empirical results reveal that while VLMs can count reliably when only one shape type is present, they exhibit substantial failures when multiple shape types are combined (i.e., compositional counting). This highlights a fundamental empirical limitation of current VLMs and motivates important directions for future research.

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。

↑