Encouraging progress has been made towards Visual Question Answering (VQA) in recent years, but it is still challenging to enable VQA models to adaptively generalize to out-of-distribution (OOD) samples. Intuitively, recompositions of existing visual concepts (i.e., attributes and objects) can generate unseen compositions in the training set, which will promote VQA models to generalize to OOD samples. In this paper, we formulate OOD generalization in VQA as a compositional generalization problem and propose a graph generative modeling-based training scheme (X-GGM) to handle the problem implicitly. X-GGM leverages graph generative modeling to iteratively generate a relation matrix and node representations for the predefined graph that utilizes attribute-object pairs as nodes. Furthermore, to alleviate the unstable training issue in graph generative modeling, we propose a gradient distribution consistency loss to constrain the data distribution with adversarial perturbations and the generated distribution. The baseline VQA model (LXMERT) trained with the X-GGM scheme achieves state-of-the-art OOD performance on two standard VQA OOD benchmarks, i.e., VQA-CP v2 and GQA-OOD. Extensive ablation studies demonstrate the effectiveness of X-GGM components.

中文翻译：

---

X-GGM：视觉问答中分布外泛化的图生成建模

近年来，视觉问答 (VQA) 取得了令人鼓舞的进展，但使 VQA 模型能够自适应地泛化到分布外 (OOD) 样本仍然具有挑战性。直观地说，现有视觉概念（即属性和对象）的重新组合可以在训练集中生成看不见的组合，这将促进 VQA 模型推广到 OOD 样本。在本文中，我们将 VQA 中的 OOD 泛化表述为组合泛化问题，并提出了一种基于图生成建模的训练方案 (X-GGM) 来隐式处理该问题。X-GGM 利用图生成建模为使用属性-对象对作为节点的预定义图迭代生成关系矩阵和节点表示。此外，为了缓解图生成建模中不稳定的训练问题，我们提出了梯度分布一致性损失来约束具有对抗性扰动的数据分布和生成的分布。使用 X-GGM 方案训练的基线 VQA 模型 (LXMERT) 在两个标准 VQA OOD 基准（即 VQA-CP v2 和 GQA-OOD）上实现了最先进的 OOD 性能。广泛的消融研究证明了 X-GGM 组件的有效性。