Standardized benchmarks like RobustBench are crucial for evaluating adversarial robustness. However, the increasing dominance of models trained on massive synthetic datasets (orders of magnitude larger than original training sets) raises questions about reported performance gains. This work identifies and investigates a potential inflation factor: high feature-level similarity between large-scale synthetic training data and benchmark test sets. We argue this similarity is an inherent characteristic arising from the probabilistic generation process of these large datasets, which naturally produces examples highly similar to test instances in feature space. This creates what we term an “Implicit Information Advantage,” where models effectively train on near-duplicates of test instances. Through comprehensive empirical analysis, we demonstrate that: (1) Synthetic datasets exhibit significantly higher similarity to the test set compared to the original training data. (2) A direct correlation exists between this similarity and robustness outcomes, with test images benefiting most having the highest similarity scores. (3) Strikingly, ablation studies show that training on just a small fraction (e.g., 1%) of the most similar synthetic examples can yield robustness comparable to using the full massive dataset. These findings suggest current benchmarks may overestimate true robust generalization due to this similarity artifact. We call for revised evaluation protocols and greater transparency to ensure benchmarks accurately measure true generalization. Code and data can be found in https://github.com/fzjcdt/RethinkingRobustBench.

Training recommendation models on large datasets is time- and resource-intensive. It is desired to construct concise yet informative datasets for efficient training. Recent advances in dataset condensation offer a promising solution by synthesizing compact datasets. However, existing methods face two key limitations when applied to recommendation: (1) they fail to generate discrete user-item interactions, and (2) they could not preserve users' potential preferences. To address the limitations, we propose a lightweight condensation framework tailored for recommendation (DConRec), focusing on condensing user-item historical interaction sets. Specifically, we model the discrete user-item interactions via a probabilistic approach and design a pre-augmentation module to incorporate the potential user preferences into the condensed datasets. While the substantial size of datasets leads to costly optimization, we propose a lightweight policy gradient estimation to accelerate the data synthesis. Experimental results on multiple real-world datasets demonstrate the effectiveness and efficiency of DConRec. Besides, we theoretically examine the provable convergence of DConRec.

Graph condensation has recently emerged as a prevalent technique to improve the training efficiency for graph neural networks (GNNs). It condenses a large graph into a small one such that a GNN trained on this small synthetic graph can achieve comparable performance to a GNN trained on the large graph. However, while existing graph condensation studies mainly focus on the best trade-off between graph size and the GNNs' performance (model utility), they overlook the security issues of graph condensation. To bridge this gap, we first explore backdoor attack against the GNNs trained on the condensed graphs. We introduce an effective backdoor attack against graph condensation, termed BGC. This attack aims to (1) preserve the condensed graph quality despite trigger injection, and (2) ensure trigger efficacy through the condensation process, achieving a high attack success rate. Specifically, BGC consistently updates triggers during condensation and targets representative nodes for poisoning. Extensive experiments demonstrate the effectiveness of our attack. BGC achieves a high attack success rate (close to 1.0) and good model utility in all cases. Furthermore, the results against multiple defense methods demonstrate BGC's resilience under their defenses. Finally, we analyze the key hyperparameters that influence the attack performance. Our code is available at: https://github.com/JiahaoWuGitIBGC.

Training recommendation models on large datasets requires significant time and resources. It is desired to construct concise yet informative datasets for efficient training. Recent advances in dataset condensation show promise in addressing this problem by synthesizing small datasets. However, applying existing methods of dataset condensation to recommendation has limitations: (1) they fail to generate discrete user-item interactions, and (2) they could not preserve users’ potential preferences. To address the limitations, we propose a lightweight condensation framework tailored for recommendation (DConRec), focusing on condensing user-item historical interaction sets. Specifically, we model the discrete user-item interactions via a probabilistic approach and design a pre-augmentation module to incorporate the potential preferences of users into the condensed datasets. While the substantial size of datasets leads to costly optimization, we propose a lightweight policy gradient estimation to accelerate the data synthesis. Experimental results on multiple real-world datasets have demonstrated the effectiveness and efficiency of our framework. Besides, we provide a theoretical analysis of the provable convergence of DConRec.

In multi-domain learning (MDL) scenarios, high labeling effort is required due to the complexity of collecting data from various domains. Active Learning (AL) presents an encouraging solution to this issue by annotating a smaller number of highly informative instances, thereby reducing the labeling effort. Previous research has relied on conventional AL strategies for MDL scenarios, which underutilize the domain-shared information of each instance during the selection procedure. To mitigate this issue, we propose a novel perturbation-based two-stage multi-domain active learning (P2S-MDAL) method incorporated into the well-regarded ASP-MTL model. Specifically, P2S-MDAL involves allocating budgets for domains and establishing regions for diversity selection, which are further used to select the most cross-domain influential samples in each region. A perturbation metric has been introduced to evaluate the robustness of the shared feature extractor of the model, facilitating the identification of potentially cross-domain influential samples. Experiments are conducted on three real-world datasets, encompassing both texts and images. The superior performance over conventional AL strategies shows the effectiveness of the proposed strategy. Additionally, an ablation study has been carried out to demonstrate the validity of each component. Finally, we outline several intriguing potential directions for future MDAL research, thus catalyzing the field's advancement.

Deep neural networks are vulnerable to adversarial examples, even in the black-box setting where the attacker is only accessible to the model output. Recent studies have devised effective black-box attacks with high query efficiency. However, such performance is often accompanied by compromises in attack imperceptibility, hindering the practical use of these approaches. In this article, we propose to restrict the perturbations to a small salient region to generate adversarial examples that can hardly be perceived. This approach is readily compatible with many existing black-box attacks and can significantly improve their imperceptibility with little degradation in attack success rates. Furthermore, we propose the Saliency Attack, a new black-box attack aiming to refine the perturbations in the salient region to achieve even better imperceptibility. Extensive experiments show that compared to the state-of-the-art black-box attacks, our approach achieves much better imperceptibility scores, including most apparent distortion (MAD), L 0 and L 2 distances, and also obtains significantly better true success rate and effective query number judged by a human-like threshold on MAD. Importantly, the perturbations generated by our approach are interpretable to some extent. Finally, it is also demonstrated to be robust to different detection-based defenses.

Social recommendations utilize social relations to enhance the representation learning for recommendations. Most social recommendation models unify user representations for the user-item interactions (collaborative domain) and social relations (social domain). However, such an approach may fail to model the users' heterogeneous behavior patterns in two domains, impairing the expressiveness of user representations. In this work, to address such limitation, we propose a novel Disentangled contrastive learning framework for social Recommendations (DcRec). More specifically, we propose to learn disentangled users' representations from the item and social domains. Moreover, disentangled contrastive learning is designed to perform knowledge transfer between disentangled users' representations for social recommendations. Comprehensive experiments on various real-world datasets demonstrate the superiority of our proposed model.

Dynamic Network Embedding (DNE) is attracting much attention due to the time-evolving nature of many real-world networks. The main objective of DNE is to efficiently update node embeddings while preserving network topology at each timestep. The idea of most existing DNE methods is to capture the topological changes at or around the most affected nodes (instead of all nodes) and accordingly update node embeddings. Unfortunately, this kind of approximation, although can improve efficiency, cannot effectively preserve the global topology of a dynamic network at each timestep, due to not considering the inactive sub-networks that receive accumulated topological changes propagated via the high-order proximity. To address this issue, we propose a new DNE method for better global topology preservation. Extensive experiments demonstrate the effectiveness and efficiency of the proposed method.

Learning low-dimensional topological representation of a network in dynamic environments is attracting much attention due to the time-evolving nature of many real-world networks. The main and common objective of Dynamic Network Embedding (DNE) is to efficiently update node embeddings while preserving network topology at each time step. The idea of most existing DNE methods is to capture the topological changes at or around the most affected nodes (instead of all nodes) and accordingly update node embeddings. Unfortunately, this kind of approximation, although can improve efficiency, cannot effectively preserve the global topology of a dynamic network at each time step, due to not considering the inactive sub-networks that receive accumulated topological changes propagated via the high-order proximity. To tackle this challenge, we propose a novel node selecting strategy to diversely select the representative nodes over a network, which is coordinated with a new incremental learning paradigm of Skip-Gram based embedding approach. The extensive experiments show GloDyNE, with a small fraction of nodes being selected, can already achieve the superior or comparable performance w.r.t. the state-of-the-art DNE methods in three typical downstream tasks. Particularly, GloDyNE significantly outperforms other methods in the graph reconstruction task, which demonstrates its ability of global topology preservation.

In real-world applications, the oracle who can label all instances correctly may not exist or may be too expensive to acquire. Alternatively, crowdsourcing provides an easy way to get labels at a low cost from multiple non-expert annotators. During the past few years, much attention has been paid to learning from such crowdsourcing data, namelyLearning from Crowds(LFC). Despite their proper statistical foundations, the existing methods for LFC still suffer from several disadvantages, such as needing prior knowledge to select the expertise model to represent the behavior of annotators, involving non-convex optimization problems, or restricting the classifier type being used. This paper addresses LFC from a quality-sensitive perspective and presents a novel framework named QS-LFC. Through reformulating the original LFC problem as a quality-sensitive learning problem, the above-mentioned disadvantages of existing methods can be avoided. Further, a support vector machine (SVM) implementation of QS-LFC is proposed. Experimental results on both synthetic and real-world data sets demonstrate that QS-LFC can achieve better generalization performance and is more robust to the noisy labels, than the existing methods.

Class evolution, the phenomenon of class emergence and disappearance, is an important research topic for data stream mining. All previous studies implicitly regard class evolution as a transient change, which is not true for many real-world problems. This paper concerns the scenario where classes emerge or disappear gradually. A class-based ensemble approach, namely Class-Based ensemble for Class Evolution (CBCE), is proposed. By maintaining a base learner for each class and dynamically updating the base learners with new data, CBCE can rapidly adjust to class evolution. A novel under-sampling method for the base learners is also proposed to handle the dynamic class-imbalance problem caused by the gradual evolution of classes. Empirical studies demonstrate the effectiveness of CBCE in various class evolution scenarios in comparison to existing class evolution adaptation methods.

Population-based Algorithm Portfolios (PAP) is an appealing framework for integrating different Evolutionary Algorithms (EAs) to solve challenging numerical optimization problems. Particularly, PAP has shown significant advantages to single EAs when a number of problems need to be solved simultaneously. Previous investigation on PAP reveals that choosing appropriate constituent algorithms is crucial to the success of PAP. However, no method has been developed for this purpose. In this paper, an extended version of PAP, namely PAP based on Estimated Performance Matrix (EPM-PAP) is proposed. EPM-PAP is equipped with a novel constituent algorithms selection module, which is based on the EPM of each candidate EAs. Empirical studies demonstrate that the EPM-based selection method can successfully identify appropriate constituent EAs, and thus EPM-PAP outperformed all single EAs considered in this work.

Hyperspectral image classification is difficult due to the high dimensional features, high intraclass variance, low interclass variance but limited training samples. In this paper, the ECASSL (Ensured Collaborative Active and Semi-Supervised Labeling) approach, which attempts to exploit pseudo-labeled samples to improve the performance of active learning based hyperspectral image classification, is proposed. In detail, in each round of active query, we obtain new human labeled samples from active query strategy and pseudo-labeled samples from the current trained classifier collaboratively. After that, we update the classifier base on latest labeled and pseudo-labeled samples. And then we correct those pseudo-labels obtained from previous iterations with the new classifier. Finally, we train the final classifier base on both the labeled samples and pseudo-labeled samples. The experiment results show that our algorithm significantly reduced the need of labeled samples while achieving comparable performance when compared with state-of-the-art algorithms for hyperspectral image classification.

Pipe breaks in urban water distribution network lead to significant economical and social costs, putting the service quality as well as the profit of water utilities at risk. To cope with such a situation, scheduled preventive maintenance is desired, which aims to predict and fix potential break pipes proactively. Physical models developed for understanding and predicting the failure of pipes are usually expensive, thus can only be used on a limited number of trunk pipes. As an alternative, statistical models that try to predict pipe breaks based on historical data are far less expensive, and therefore have attracted a lot of interests from water utilities recently. In this paper, we report a novel data mining prediction system that has been built for a water utility in a big Chinese city. Various aspects of how to build such a system are described, including problem formulation, data cleaning, model construction, as well as evaluating the importance of attributes according to the requirements of end users in water utilities. Satisfactory results have been achieved by our prediction system. For example, with the system trained on the available dataset at the end of 2010, the water utility would avoid 50% of pipe breaks in 2011 by examining only 6.98% of its pipes in advance. During the construction of the system, we find that the extremely skew distribution of break and non-break pipes, interestingly, is not an obstacle. This lesson could serve as a practical reference for both academical studies on imbalanced learning as well as future explorations on pipe failure prediction problems.