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 2410.01690cs.AI
收藏

为什么情境在 VQA 与推理中至关重要:面向 VLM 输入模态的语义干预

Why context matters in VQA and Reasoning: Semantic interventions for VLM input modalities

ETH Zurich(苏黎世联邦理工学院);ETH AI Center(ETH人工智能中心);German Cancer Research Center (DKFZ)(德国癌症研究中心);Helmholtz Imaging(亥姆霍兹成像中心);IBM Research(IBM研究院)

Kenza Amara, Lukas Klein, Carsten Lüth, Paul Jäger, Hendrik Strobelt, Mennatallah El-Assady

AI总结:

提出 SI-VQA 数据集与 ISI 工具,通过语义干预系统评估图像和文本模态对 VLM 在 VQA 与推理中准确率、不确定性及注意力的影响。

摘要:

生成式 AI 的各种局限(如幻觉和模型失效)使得理解不同模态在视觉语言模型(VLM)预测中的作用变得至关重要。我们的工作研究图像与文本模态信息的融合如何影响 VLM 在视觉问答(VQA)和推理任务中的性能与行为。我们通过答案准确率、推理质量、模型不确定性和模态相关性来衡量这一影响。我们在视觉内容对解决 VQA 任务必不可少的不同配置下,研究文本与图像模态之间的相互作用。我们的贡献包括:(1)Semantic Interventions (SI)-VQA 数据集;(2)在不同模态配置下对多种 VLM 架构开展的基准研究;以及(3)Interactive Semantic Interventions (ISI) 工具。SI-VQA 数据集是该基准的基础,而 ISI 工具提供了一个界面,用于测试并应用对图像和文本输入的语义干预,从而实现更细粒度的分析。结果表明,模态间的互补信息可提升答案与推理质量,而矛盾信息会损害模型性能和置信度。图像文本标注对准确率和不确定性影响极小,并会略微提高图像相关性。注意力分析证实,在 VQA 任务中图像输入相对于文本占据主导地位。本研究评估了若干允许提取各模态注意力系数的最先进 VLM。一个关键发现是 PaliGemma 存在有害的过度自信问题;与 LLaVA 模型相比,这会带来更高的静默失效风险。本工作为严格分析模态融合奠定了基础,并由专门为此设计的数据集提供支持。

英文摘要:

The various limitations of Generative AI, such as hallucinations and model failures, have made it crucial to understand the role of different modalities in Visual Language Model (VLM) predictions. Our work investigates how the integration of information from image and text modalities influences the performance and behavior of VLMs in visual question answering (VQA) and reasoning tasks. We measure this effect through answer accuracy, reasoning quality, model uncertainty, and modality relevance. We study the interplay between text and image modalities in different configurations where visual content is essential for solving the VQA task. Our contributions include (1) the Semantic Interventions (SI)-VQA dataset, (2) a benchmark study of various VLM architectures under different modality configurations, and (3) the Interactive Semantic Interventions (ISI) tool. The SI-VQA dataset serves as the foundation for the benchmark, while the ISI tool provides an interface to test and apply semantic interventions in image and text inputs, enabling more fine-grained analysis. Our results show that complementary information between modalities improves answer and reasoning quality, while contradictory information harms model performance and confidence. Image text annotations have minimal impact on accuracy and uncertainty, slightly increasing image relevance. Attention analysis confirms the dominant role of image inputs over text in VQA tasks. In this study, we evaluate state-of-the-art VLMs that allow us to extract attention coefficients for each modality. A key finding is PaliGemma's harmful overconfidence, which poses a higher risk of silent failures compared to the LLaVA models. This work sets the foundation for rigorous analysis of modality integration, supported by datasets specifically designed for this purpose.

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。

↑