Attribution explanation is a typical approach for interpreting deep neural networks (DNNs), aiming to quantify the contribution score of individual input variables to model predictions. Despite extensive methodological development, a fundamental faithfulness problem remains unresolved: whether existing attribution methods faithfully reflect the true decision-making logic of DNNs, which significantly limits their reliability and practical adoption. These concerns largely stem from three core challenges: the lack of a unified theoretical framework, clear theoretical rationales, and principled faithfulness evaluation in the absence of ground truth. Recently, a growing body of theoretical studies has begun to address these issues, marking an important shift toward principled understanding of attribution methods. In this survey, we provide a comprehensive review of these advances, with a particular emphasis on three interconnected directions: (i) Theoretical unification, which uncovers key commonalities and differences among attribution methods; (ii) Theoretical rationale, which clarifies the mathematical and conceptual justifications underlying existing methods; (iii) Theoretical evaluation, which rigorously proves whether attribution methods satisfy established faithfulness principles. Beyond a comprehensive review, we provide practical recommendations and a case study illustrating how theoretical findings can be translated into operational decision rules for method design, selection, and usage. We conclude with a discussion of promising open problems for further work.

Large Language Models (LLMs) may suffer from hallucinations in real-world applications due to the lack of relevant knowledge. In contrast, knowledge graphs encompass extensive, multi-relational structures that store a vast array of symbolic facts. Consequently, integrating LLMs with knowledge graphs has been extensively explored, with Knowledge Graph Question Answering (KGQA) serving as a critical touchstone for the integration. This task requires LLMs to answer natural language questions by retrieving relevant triples from knowledge graphs. However, existing methods face two significant challenges: *excessively long reasoning paths distracting from the answer generation*, and *false-positive relations hindering the path refinement*. In this paper, we propose an iterative interactive KGQA framework that leverages the interactive learning capabilities of LLMs to perform reasoning and Debating over Graphs (DoG). Specifically, DoG employs a subgraph-focusing mechanism, allowing LLMs to perform answer trying after each reasoning step, thereby mitigating the impact of lengthy reasoning paths. On the other hand, DoG utilizes a multi-role debate team to gradually simplify complex questions, reducing the influence of false-positive relations. This debate mechanism ensures the reliability of the reasoning process. Experimental results on five public datasets demonstrate the effectiveness and superiority of our architecture. Notably, DoG outperforms the state-of-the-art method ToG by 23.7% and 9.1% in accuracy on WebQuestions and GrailQA, respectively. Furthermore, the integration experiments with various LLMs on the mentioned datasets highlight the flexibility of DoG.

This paper introduces TexGarment, an efficient method for synthesizing high-quality, 3D-consistent garment textures in UV space. Traditional approaches based on 2D-to-3D mapping often suffer from 3D inconsistency, while methods learning from limited 3D data lack sufficient texture diversity. These limitations are particularly problematic in garment texture generation, where high demands exist for both detail and variety. To address these challenges, TexGarment leverages a pre-trained text-to-image diffusion Transformer model with robust generalization capabilities, introducing structural information to guide the model in generating 3D-consistent garment textures in a single inference step. Specifically, We utilize the 2D UV position map to guide the layout during the UV texture generation process, ensuring a coherent texture arrangement and enhancing it by integrating global 3D structural information from the mesh surface point cloud. This combined guidance effectively aligns 3D structural integrity with 2D layout. Our method efficiently generates high-quality, diverse UV textures in a single inference step while maintaining 3D consistency. Experimental results validate the effectiveness of TexGarment, achieving state-of-the-art performance in 3D garment texture generation.

Individual trajectory prediction plays a crucial role in intelligent transportation systems. While existing methods demonstrate strong performance in short-term forecasting (e.g., minute-level predictions), they are limited in modeling long-term patterns (day-level predictions). The key challenge is capturing both the periodic regularity and stochastic variability of urban mobility. To bridge this gap, we propose VF-Diffusion, a novel framework for long-term individual trajectory prediction with three key innovations: (1) A direction-sensitive diffusion model that generates baseline trajectories by learning motion trends; (2) A trajectory rectification module that refines spatial displacements using historical median coordinates; and (3) A frequency-sensitive mechanism that identifies high-frequency visit locations, predicts their temporal sequences via an ensemble model, and integrates them with the baseline trajectory. By combining generative modeling with a frequency-sensitive mechanism, VF-Diffusion fills a critical gap in existing methods, offering the ability to predict new visiting areas and improve trajectory accuracy. Extensive experiments on Beijing Wi-Fi trajectory data show that our method outperforms four baselines, achieving about 90% accuracy for predictions within a 1 km threshold. It particularly excels in areas with frequent and periodic visits. This framework advances trajectory prediction by enabling multi-day forecasting, a previously underexplored capability, and offers practical solutions for enhancing smart city infrastructure.

Pseudo-labeling is a widely used strategy in semi-supervised learning. Existing methods typically select predicted labels with high confidence scores and high training stationarity, as pseudo-labels to augment training sets. In contrast, this paper explores the pseudo-labeling potential of predicted labels that **do not** exhibit these characteristics. We discover a new type of predicted labels suitable for pseudo-labeling, termed *two-phase labels*, which exhibit a two-phase pattern during training: *they are initially predicted as one category in early training stages and switch to another category in subsequent epochs.* Case studies show the two-phase labels are informative for decision boundaries. To effectively identify the two-phase labels, we design a 2-*phasic* metric that mathematically characterizes their spatial and temporal patterns. Furthermore, we propose a loss function tailored for two-phase pseudo-labeling learning, allowing models not only to learn correct correlations but also to eliminate false ones. Extensive experiments on eight datasets show that **our proposed 2-*phasic* metric acts as a powerful booster** for existing pseudo-labeling methods by additionally incorporating the two-phase labels, achieving an average classification accuracy gain of 1.73% on image datasets and 1.92% on graph datasets.

Federated learning (FL) is a popular collaborative training paradigm in which data owners offer gradients instead of private data to model owners for model training to protect data privacy. However, it faces security threats from two sides: dishonest model owners may extract sensitive information about private data from gradients; meanwhile, adversaries may pretend to be data owners and poison the model by sending malicious gradients. We propose a novel FL protocol, PARSIFAL, to address privacy leakage and model poisoning threats. A poisoning detection module is designed based on a novel sketch structure. This module efficiently detects potential malicious gradients that are dissimilar to the majority of benign gradients. PARSIFAL also contains a robust aggregation module based on sign gradients to mitigate the influence of poisoning gradients on aggregation results. Meanwhile, all processes of our PARSIFAL are protected by privacy protocols, mainly based on secret sharing, to guarantee that malicious detection and aggregation processes will not leak sensitive information. Experimental results show that PARSIFAL improves poisoning defense performance by up to 28% compared with recent baselines.

Knowledge graph-based retrieval-augmented generation seeks to mitigate hallucinations in Large Language Models (LLMs) caused by insufficient or outdated knowledge. However, existing methods often fail to fully exploit the prior knowledge embedded in knowledge graphs (KGs), particularly their structural information and explicit or implicit constraints. The former can enhance the faithfulness of LLMs' reasoning, while the latter can improve the reliability of response generations. Motivated by these, we propose a trustworthy reasoning framework, termed Deliberation over Priors (\texttt{DP}), which sufficiently utilizes the priors contained in KGs. Specifically, \texttt{DP} adopts a progressive knowledge distillation strategy that integrates structural priors into LLMs through a combination of supervised fine-tuning and Kahneman-Tversky Optimization, thereby improving the faithfulness of relation path generation. Furthermore, our framework employs a reasoning-introspection strategy, which guides LLMs to perform refined reasoning verification based on extracted constraint priors, ensuring the reliability of response generation. Extensive experiments on three benchmark datasets demonstrate that \texttt{DP} achieves new state-of-the-art performance, especially a H@1 improvement of 13% on the ComplexWebQuestions dataset, and generates highly trustworthy responses. We also conduct various analyses to verify its flexibility and practicality. Code is available at [https://github.com/mira-ai-lab/Deliberation-on-Priors](https://github.com/mira-ai-lab/Deliberation-on-Priors).

Variational Inference (VI) has gained popularity as a flexible approximate inference scheme for computing posterior distributions in Bayesian models. Original VI methods use Kullback-Leibler (KL) divergence to construct variational objectives. However, KL divergence has zero-forcing behavior and is completely agnostic to the metric of the underlying data distribution, resulting in bad approximations. To alleviate this issue, we propose a new variational objective by using Optimal Transport (OT) distance, which is a metric-aware divergence, to measure the difference between approximate posteriors and priors. The superior performance of OT distance enables us to learn more accurate approximations. We further enhance the objective by gradually including the OT term using a hyperparameter λ for over-parameterized models. We develop a Variational inference method with OT (VOT) which presents a gradient-based black-box framework for solving Bayesian models, even when the density function of approximate distribution is not available. We provide the consistency analysis of approximate posteriors and demonstrate the practical effectiveness on Bayesian neural networks and variational autoencoders.

Molecular representation learning has emerged as a game-changer at the intersection of AI and chemistry, with great potential in applications such as drug design and materials discovery. A substantial obstacle in successfully applying molecular representation learning is the difficulty of effectively and completely characterizing and learning molecular geometry, which has not been well addressed to date. To overcome this challenge, we propose a novel framework that features a novel geometric graph, termed HAGO-Graph, and a specifically designed geometric graph learning model, HAGO-Net. In the framework, the foundation is HAGO-Graph, which enables a complete characterization of molecular geometry in a hierarchical manner. Specifically, we leverage the concept of n-body in physics to characterize geometric patterns at multiple spatial scales. We then specifically design a message passing scheme, HAGO-MPS, and implement the scheme as a geometric graph neural network, HAGO-Net, to effectively learn the representation of HAGO-Graph by horizontal and vertical aggregation. We further prove DHAGO-Net, the derivative function of HAGO-Net, is an equivariant model. The proposed models are validated by extensive comparisons on four challenging benchmarks. Notably, the models exhibited state-of-the-art performance in molecular chirality identification and property prediction, achieving state-of-the-art performance on five properties of QM9 dataset. The models also achieved competitive results on molecular dynamics prediction task.

The advancement toward deeper graph neural networks is currently obscured by two inherent issues in message passing, oversmoothing and oversquashing. We identify the root cause of these issues as information loss due to heterophily mixing in aggregation, where messages of diverse category semantics are mixed. We propose a novel multi-track graph convolutional network to address oversmoothing and oversquashing effectively. Our basic idea is intuitive: if messages are separated and independently propagated according to their category semantics, heterophilic mixing can be prevented. Consequently, we present a novel multi-track message passing scheme capable of preventing heterophilic mixing, enhancing long-distance information flow, and improving separation condition. Empirical validations show that our model achieved state-of-the-art performance on several graph datasets and effectively tackled oversmoothing and oversquashing, setting a new benchmark of $86.4$% accuracy on Cora.

Audio-Visual Question Answering (AVQA) is a complex multi-modal reasoning task, demanding intelligent systems to accurately respond to natural language queries based on audio-video input pairs. Nevertheless, prevalent AVQA approaches are prone to overlearning dataset biases, resulting in poor robustness. Furthermore, current datasets may not provide a precise diagnostic for these methods. To tackle these challenges, firstly, we propose a novel dataset, *MUSIC-AVQA-R*, crafted in two steps: rephrasing questions within the test split of a public dataset (*MUSIC-AVQA*) and subsequently introducing distribution shifts to split questions. The former leads to a large, diverse test space, while the latter results in a comprehensive robustness evaluation on rare, frequent, and overall questions. Secondly, we propose a robust architecture that utilizes a multifaceted cycle collaborative debiasing strategy to overcome bias learning. Experimental results show that this architecture achieves state-of-the-art performance on MUSIC-AVQA-R, notably obtaining a significant improvement of 9.32\%. Extensive ablation experiments are conducted on the two datasets mentioned to analyze the component effectiveness within the debiasing strategy. Additionally, we highlight the limited robustness of existing multi-modal QA methods through the evaluation on our dataset. We also conduct experiments combining various baselines with our proposed strategy on two datasets to verify its plug-and-play capability. Our dataset and code are available at <https://github.com/reml-group/MUSIC-AVQA-R>.

Visual question answering requires a system to provide an accurate natural language answer given an image and a natural language question. However, it is widely recognized that previous generic VQA methods often tend to memorize biases present in the training data rather than learning proper behaviors, such as grounding images before predicting answers. Therefore, these methods usually achieve high in-distribution but poor out-of-distribution performance. In recent years, various datasets and debiasing methods have been proposed to evaluate and enhance the VQA robustness, respectively. This paper provides the first comprehensive survey focused on this emerging fashion. Specifically, we first provide an overview of the development process of datasets from in-distribution and out-of-distribution perspectives. Then, we examine the evaluation metrics employed by these datasets. Third, we propose a typology that presents the development process, similarities and differences, robustness comparison, and technical features of existing debiasing methods. Furthermore, we analyze and discuss the robustness of representative vision-and-language pre-training models on VQA. Finally, through a thorough review of the available literature and experimental analysis, we discuss the key areas for future research from various viewpoints.

This paper tackles the cross-modality person re-identification (re-ID) problem by suppressing the modality discrepancy. In cross-modality re-ID, the query and gallery images are in different modalities. Given a training identity, the popular deep classification baseline shares the same proxy (i.e., a weight vector in the last classification layer) for two modalities. We find that it has considerable tolerance for the modality gap, because the shared proxy acts as an intermediate relay between two modalities. In response, we propose a Memory-Augmented Unidirectional Metric (MAUM) learning method consisting of two novel designs, i.e., unidirectional metrics, and memory-based augmentation. Specifically, MAUM first learns modality-specific proxies (MS-Proxies) independently under each modality. Afterward, MAUM uses the already-learned MS-Proxies as the static references for pulling close the features in the counterpart modality. These two unidirectional metrics (IR image to RGB proxy and RGB image to IR proxy) jointly alleviate the relay effect and benefit cross-modality association. The cross-modality association is further enhanced by storing the MS-Proxies into memory banks to increase the reference diversity. Importantly, we show that MAUM improves cross-modality re-ID under the modality-balanced setting and gains extra robustness against the modality-imbalance problem. Extensive experiments on SYSU-MMOI and RegDB datasets demonstrate the superiority of MAUM over the state-of-the-art. The code will be available.

The key to the effective control of a diffusion system lies in how accurately we could predict its unfolding dynamics based on the observation of its current state. However, in the real-world applications, it is often infeasible to conduct a timely and yet comprehensive observation due to resource constraints. In view of such a practical challenge, the goal of this work is to develop a novel computational method for performing active observations, termed active surveillance, with limited resources. Specifically, we aim to predict the dynamics of a large spatio-temporal diffusion system based on the observations of some of its components. Towards this end, we introduce a novel measure, the γ value, that enables us to identify the key components by means of modeling a sentinel network with a row sparsity structure. Having obtained a theoretical understanding of the γ value, we design a backward-selection sentinel network mining algorithm (SNMA) for deriving the sentinel network via group sparse Bayesian learning. In order to be practically useful, we further address the issue of scalability in the computation of SNMA, and moreover, extend SNMA to the case of a non-linear dynamical system that could involve complex diffusion mechanisms. We show the effectiveness of SNMA by validating it using both synthetic datasets and five real-world datasets. The experimental results are appealing, which demonstrate that SNMA readily outperforms the state-of-the-art methods.

This paper considers deep visual recognition on long-tailed data. Under the long-tailed distribution, a small portion of the classes (head classes) occupy most training samples and the most classes (tail classes) only occupy relatively few samples. We observe that such long-tailed distribution significantly distorts the deeply-learned feature space, which consequentially compromises the deep visual recognition. Specifically, during training, each head class is prone to a relatively wide spatial distribution in the deep feature space, while each tail class is prone to a relatively small spatial distribution. In another word, the tail classes usually have much smaller spatial distribution than the head classes, distorting the overall feature space. In response, we propose to explicitly inflate the distribution of each tail class in the deep feature space, so that the tail classes will have comparable distribution range as the head classes. To this end, we replace each tail feature vector with a set of feature vectors on the fly. These feature vectors follow a probabilistic distribution learned from the head classes and yield a “feature cloud” surrounding the original tail feature. We show that the feature cloud effectively transfers the within-class diversity from the head classes onto the tail classes, maintaining an effect of probabilistic feature augmentation. An important advantage of the proposed feature cloud is that it is capable to bring general improvement to long-tailed visual recognition on two fundamental tasks,i.e., deep classification and deep representation learning, in spite of the significant differences between them. Extensive experiments on both deep metric learning benchmarks and deep image classification benchmarks validate the effectiveness of the proposed feature cloud.

Recommender systems aim to accurately and actively provide users with potentially interesting items (products, information or services). Deep reinforcement learning has been successfully applied to recommender systems, but still heavily suffer from data sparsity and cold-start in real-world tasks. In this work, we propose an effective way to address such issues by leveraging the pervasive social networks among users in the estimation of action-values (Q). Specifically, we develop a Social Attentive Deep Q-network (SADQN) to approximate the optimal action-value function based on the preferences of both individual users and social neighbors, by successfully utilizing a social attention layer to model the influence between them. Further, we propose an enhanced variant of SADQN, termed SADQN++, to model the complicated and diverse trade-offs between personal preferences and social influence for all involved users, making the agent more powerful and flexible in learning the optimal policies. The experimental results on real-world datasets demonstrate that the proposed SADQNs remarkably outperform the state-of-the-art deep reinforcement learning agents, with reasonable computation cost.

Recent research on graph embedding has achieved success in various applications. Most graph embedding methods preserve the proximity in a graph into a manifold in an embedding space. We argue an important but neglected problem about this proximity-preserving strategy: Graph topology patterns, while preserved well into an embedding manifold by preserving proximity, may distort in the ambient embedding Euclidean space, and hence to detect them becomes difficult for machine learning models. To address the problem, we propose curvature regularization, to enforce flatness for embedding manifolds, thereby preventing the distortion. We present a novel angle-based sectional curvature, termed ABS curvature, and accordingly three kinds of curvature regularization to induce flat embedding manifolds during graph embedding. We integrate curvature regularization into five popular proximity-preserving embedding methods, and empirical results in two applications show significant improvements on a wide range of open graph datasets.

Reinforcement learning (RL) has been successfully applied to recommender systems. However, the existing RL-based recommendation methods are limited by their unstructured state/action representations. To address this limitation, we propose a novel way that builds high-quality graph-structured states/actions according to the user-item bipartite graph. More specifically, we develop an end-to-end RL agent, termed Graph Convolutional Q-network (GCQN), which is able to learn effective recommendation policies based on the inputs of the proposed graph-structured representations. We show that GCQN achieves significant performance margins over the existing methods, across different datasets and task settings.

While deep reinforcement learning has been successfully applied to recommender systems, it is challenging and unexplored to improve the performance of deep reinforcement learning recommenders by effectively utilizing the pervasive social networks. In this work, we develop a Social Attentive Deep Q-network (SADQN) agent, which is able to provide high-quality recommendations during user-agent interactions by leveraging social influence among users. Specifically, SADQN is able to estimate action-values not only based on the users' personal preferences, but also based on their social neighbors' preferences by employing a particular social attention layer. The experimental results on three real-world datasets demonstrate that SADQN significantly improves the performance of deep reinforcement learning agents that overlook social influence.

Predicting epidemic dynamics is of great value in understanding and controlling diffusion processes, such as infectious disease spread and information propagation. This task is intractable, especially when surveillance resources are very limited. To address the challenge, we study the problem of active surveillance, i.e., how to identify a small portion of system components as sentinels to effect monitoring, such that the epidemic dynamics of an entire system can be readily predicted from the partial data collected by such sentinels. We propose a novel measure, the gamma value, to identify the sentinels by modeling a sentinel network with row sparsity structure. We design a flexible group sparse Bayesian learning algorithm to mine the sentinel network suitable for handling both linear and non-linear dynamical systems by using the expectation maximization method and variational approximation. The efficacy of the proposed algorithm is theoretically analyzed and empirically validated using both synthetic and real-world data.

During an epidemic, the spatial, temporal and demographic patterns of disease transmission are determined by multiple factors. In addition to the physiological properties of the pathogens and hosts, the social contact of the host population, which characterizes the reciprocal exposures of individuals to infection according to their demographic structure and various social activities, are also pivotal to understanding and predicting the prevalence of infectious diseases. How social contact is measured will affect the extent to which we can forecast the dynamics of infections in the real world. Most current work focuses on modeling the spatial patterns of static social contact. In this work, we use a novel perspective to address the problem of how to characterize and measure dynamic social contact during an epidemic. We propose an epidemic-model-based tensor deconvolution framework in which the spatiotemporal patterns of social contact are represented by the factors of the tensors. These factors can be discovered using a tensor deconvolution procedure with the integration of epidemic models based on rich types of data, mainly heterogeneous outbreak surveillance data, socio-demographic census data and physiological data from medical reports. Using reproduction models that include SIR/SIS/SEIR/SEIS models as case studies, the efficacy and applications of the proposed framework are theoretically analyzed, empirically validated and demonstrated through a set of rigorous experiments using both synthetic and real-world data.

During an epidemic, the spatial, temporal and demographical patterns of disease transmission are determined by multiple factors. Besides the physiological properties of pathogenes and hosts, the social contacts of host population, which characterize individuals' reciprocal exposures of infection in view of demographical structures and various social activities, are also pivotal to understand and further predict the prevalence of infectious diseases. The means of measuring social contacts will dominate the extent how precisely we can forecast the dynamics of infections in the real world. Most current works focus their efforts on modeling the spatial patterns of static social contacts. In this work, we address the problem on how to characterize and measure dynamical social contacts during an epidemic from a novel perspective. We propose an epidemic-model-based tensor deconvolution framework to address this issue, in which the spatiotemporal patterns of social contacts are represented by the factors of tensors, which can be discovered by a tensor deconvolution procedure with an integration of epidemic models from rich types of data, mainly including heterogeneous outbreak surveillance, social-demographic census and physiological data from medical reports. Taking SIR model as a case study, the efficacy of the proposed method is theoretically analyzed and empirically validated through a set of rigorous experiments on both synthetic and real-world data.

In this paper, a method about how to identify insecure behaviors of browser extensions is proposed. Typically, the identification of insecure extension behaviors is based on knowledge which is got by investigating known malicious or vulnerable extensions. We present an automatic technique that can ease the laborious manual investigating process. Our technique mines the difference between the behavior graphs of insecure and secure extensions based on graph mining algorithm. The difference between them is the specification of insecure extension behaviors which can be further analyzed manually or automatically to help people make a better decision about whether an extension is secure or not. We developed a prototype and the experimental results show that this kind of technique can effectively extract insecure extension behaviors.