Siheng Li, Kejiao Li, Zenan Xu, Guanhua Huang, Kun Li, Haoyuan Wu, Wujiajia, Zihao Zheng, Chenchen Zhang, Kun Shi, Xue Gong, Qi Yi, Ruibin Xiong, Tingqiang Xu, Yuhao Jiang, Jianfeng Yan, Yuyuan Zeng, Guanghui Xu, Jinbao Xue, Zhijiang xu, Zheng Fang, Shuai LI, Qibin Liu, Xiaoxue Li, Zhuoyu Li, Yangyu Tao, Fei Gao, Cheng Jiang, Bochao Wang, Kai Liu, Jianchen Zhu, Wai Lam, Bo Zhou, Di Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Deep models have been widely and successfully used in image manipulation detection, which aims to classify tampered images and localize tampered regions. Most existing methods mainly focus on extracting global features from tampered images, while neglecting the relationships of local features between tampered and authentic regions within a single tampered image. To exploit such spatial relationships, we propose Proposal Contrastive Learning (PCL) for effective image manipulation detection. Our PCL consists of a two-stream architecture by extracting two types of global features from RGB and noise views respectively. To further improve the discriminative power, we exploit the relationships of local features through a proxy proposal contrastive learning task by attracting/repelling proposal-based positive/negative sample pairs. Moreover, we show that our PCL can be easily adapted to unlabeled data in practice, which can reduce manual labeling costs and promote more generalizable features. Extensive experiments among several standard datasets demonstrate that our PCL can be a general module to obtain consistent improvement. The code is available athttps://github.com/Sandy-Zeng/PCL.

This paper concerns the problem of network embedding (NE), which aims to learn low-dimensional representations for network nodes. Such dense representations offer great promises for many network analysis problems. However, existing approaches are still faced with challenges posed by the characteristics of complex real-world networks. First, for networks associated with rich content information, previous methods often learn separated content and structure representations, which requires post-processing of combination. Empirical combination strategies often make the final vectors suboptimal. Second, existing methods preserve the structure information by considering short and fixed neighborhood scope, such as the first- and/or the second-order proximities. However, it is hard to decide the neighborhood scope in complex problems. To this end, we propose a novel sequence to sequence model based NE framework referred to as Self-Translation Network Embedding (STNE). With the sampled node sequences, STNE translates each sequence itself from the content sequence to the node sequence. On the one hand, the bi-directional LSTM encoder fuses the content and structure information seamlessly from the raw input. On the other hand, high-order proximity can be flexibly learned with the memories of LSTM to capture long-range structural information. Experimental results on three real-world datasets demonstrate the superiority of STNE.

Convolutional neural networks (CNNs) can be easily over-fitted when they are over-parametered. The popular dropout that drops feature units randomly can't always work well for CNNs, due to the problem of under-dropping. To eliminate this problem, some structural dropout methods such as SpatialDropout, Cutout and DropBlock have been proposed. However, these methods that drop feature units in continuous regions randomly, may have the risk of over-dropping, thus leading to degradation of performance. To address these issues, we propose a novel structural dropout method, Correlation based Dropout (CorrDrop), to regularize CNNs by dropping feature units based on feature correlation, which reflects the discriminative information in feature maps. Specifically, the proposed method first obtains correlation map based on the activation in the feature maps, and then adaptively masks out those regions with small average correlation. Thus, the proposed method can regularize CNNs well by discarding part of contextual regions. Extensive experiments on image classification demonstrate the superiority of our method compared with other counterparts.