--Audio-Visual Question Answering (A VQA) is a challenging multimodal reasoning task requiring intelligent systems to answer natural language queries based on paired audio-video inputs accurately. However, existing A VQA approaches often suffer from overfitting to dataset biases, leading to poor robustness. T o address these challenges, we first introduce a novel dataset, FortisA VQA, constructed in two stages: (1) rephrasing questions in the test split of the public MUSIC-A VQA dataset and (2) introducing distribution shifts across questions. The first stage expands the test space with greater diversity, while the second enables a refined robustness evaluation across rare, frequent, and overall question distributions. Second, we introduce a robust Multimodal Audio-Visual Epistemic Network (MA VEN) that leverages a multifaceted cycle collaborative debiasing strategy to mitigate bias learning. Experimental results demonstrate that our architecture achieves state-of-the-art performance on FortisA VQA, with a notable improvement of 7.81%. Additionally, our evaluation reveals the limited robustness of existing multimodal QA methods. We also verify the plug-and-play capability of our strategy by integrating it with various baseline models across both datasets. UMANS possess the extraordinary capacity to seam-lessly integrate auditory and visual cues, effectively establishing a cohesive relationship between visual and auditory stimuli [1-3]. Jie Ma, Pinghui Wang, Jing Tao and Zhou Su are with the Ministry of Education of Key Laboratory for Intelligent Networks and Network Security, School of Cyber Science and Engineering, Xi'an Jiaotong University, Xi'an, Shaanxi 710049, China. Zhitao Gao and Jun Liu are with the Shannxi Provincial Key Laboratory of Big Data Knowledge Engineering, School of Computer Science and Technology, Xi'an Jiaotong University, Xi'an, Shaanxi 710049, China. Qi Chai is with the Information Hub, Hong Kong University of Science and Technology (Guangzhou), Guangzhou, Guangdong, 510000, China. The question in current A VQA datasets is generated by a limited set of predefined templates, which may not be in line with the real-world scenario. Our findings indicate that existing methods such as STG [6] are not robust, which may be attributed to excessive bias learning, such as memorizing statistical regularities between critical question words and answers. It requires the system to learn high-order interaction representations of the concepts encompassed with audio, video, and language modalities. As is known to us [8-10], the high-level reasoning ability of the system mainly relies on large-scale data that does not contain harmful biases or statistical regularities. However, completely avoiding the negative bias in datasets seems challenging [11] due to the inherent skewness in real-world data distributions.

With the continuous evolution of AI-generated images, the generalized detection of them has become a crucial aspect of AI security. Existing detectors have focused on cross-generator generalization, while it remains unexplored whether these detectors can generalize across different image scenes, e.g., images from different datasets with different semantics. In this paper, we reveal that existing detectors suffer from substantial accuracy drops in such cross-scene generalization. In particular, we attribute their failures to "semantic artifacts" in both real and generated images, to which detectors may overfit. To break such "semantic artifacts", we propose a simple yet effective approach based on conducting an image patch shuffle and then training an end-to-end patch-based classifier. We conduct a comprehensive openworld evaluation on 31 test sets, covering 7 Generative Adversarial Networks, 18 (variants of) Diffusion Models, and another 6 CNN-based generative models. The results demonstrate that our approach outperforms previous approaches by 2.08% (absolute) on average regarding cross-scene detection accuracy. We also notice the superiority of our approach in open-world generalization, with an average accuracy improvement of 10.59% (absolute) across all test sets.

Visual Grounding aims to localize the referring object in an image given a natural language expression. Recent advancements in DETR-based visual grounding methods have attracted considerable attention, as they directly predict the coordinates of the target object without relying on additional efforts, such as pre-generated proposal candidates or pre-defined anchor boxes. However, existing research primarily focuses on designing stronger multi-modal decoder, which typically generates learnable queries by random initialization or by using linguistic embeddings. This vanilla query generation approach inevitably increases the learning difficulty for the model, as it does not involve any target-related information at the beginning of decoding. Furthermore, they only use the deepest image feature during the query learning process, overlooking the importance of features from other levels.

Point-based methods have made significant progress, but improving their scalability in large-scale 3D scenes is still a challenging problem. In this paper, we delve into the point-based method and develop a simpler, faster, stronger variant model, dubbed as LinNet. In particular, we first propose the disassembled set abstraction (DSA) module, which is more effective than the previous version of set abstraction. It achieves more efficient local aggregation by leveraging spatial anisotropy and channel anisotropy separately. Additionally, by mapping 3D point clouds onto 1D space-filling curves, we enable parallelization of downsampling and neighborhood queries on GPUs with linear complexity. LinNet, as a purely point-based method, outperforms most previous methods in both indoor and outdoor scenes without any extra attention, and sparse convolution but merely relying on a simple MLP. It achieves the mIoU of 73.7%, 81.4%, and 69.1% on the S3DIS Area5, NuScenes, and SemanticKITTI validation benchmarks, respectively, while speeding up almost 10x times over PointNeXt. Our work further reveals both the efficacy and efficiency potential of the vanilla pointbased models in large-scale representation learning. Our code will be available at https://github.com/DengH293/LinNet.

Existing prompt-tuning methods have demonstrated impressive performances in continual learning (CL), by selecting and updating relevant prompts in the visiontransformer models. On the contrary, this paper aims to learn each task by tuning the prompts in the direction orthogonal to the subspace spanned by previous tasks' features, so as to ensure no interference on tasks that have been learned to overcome catastrophic forgetting in CL. However, different from the orthogonal projection in the traditional CNN architecture, the prompt gradient orthogonal projection in the ViT architecture shows completely different and greater challenges, i.e., 1) the highorder and non-linear self-attention operation; 2) the drift of prompt distribution brought by the LayerNorm in the transformer block. Theoretically, we have finally deduced two consistency conditions to achieve the prompt gradient orthogonal projection, which provide a theoretical guarantee of eliminating interference on previously learned knowledge via the self-attention mechanism in visual prompt tuning. In practice, an effective null-space-based approximation solution has been proposed to implement the prompt gradient orthogonal projection. Extensive experimental results demonstrate the effectiveness of anti-forgetting on four classincremental benchmarks with diverse pre-trained baseline models, and our approach achieves superior performances to state-of-the-art methods.

Recent advances in prompt optimization have notably enhanced the performance of pre-trained language models (PLMs) on downstream tasks. However, the potential of optimized prompts on domain generalization has been under-explored. To explore the nature of prompt generalization on unknown domains, we conduct pilot experiments and find that (i) Prompts gaining more attention weight from PLMs' deep layers are more generalizable and (ii) Prompts with more stable attention distributions in PLMs' deep layers are more generalizable. Thus, we offer a fresh objective towards domain-generalizable prompts optimization named "Concentration", which represents the "lookback" attention from the current decoding token to the prompt tokens, to increase the attention strength on prompts and reduce the fluctuation of attention distribution. We adapt this new objective to popular soft prompt and hard prompt optimization methods, respectively. Extensive experiments demonstrate that our idea improves comparison prompt optimization methods by 1.42% for soft prompt generalization and 2.16% for hard prompt generalization in accuracy on the multi-source domain generalization setting, while maintaining satisfying in-domain performance. The promising results validate the effectiveness of our proposed prompt optimization objective and provide key insights into domain-generalizable prompts.

The construction of Generalized Knowledge Graph (GKG), including knowledge graph, event knowledge graph and commonsense knowledge graph, is fundamental for various natural language processing tasks. Current studies typically construct these types of graph separately, overlooking holistic insights and potential unification that could be beneficial in computing resources and usage perspectives. However, a key challenge in developing a unified framework for GKG is obstacles arising from task-specific differences. In this study, we propose a unified framework for constructing generalized knowledge graphs to address this challenge. First, we collect data from 15 sub-tasks in 29 datasets across the three types of graphs, categorizing them into in-sample, counter-task, and out-of-distribution (OOD) data. Then, we propose a three-stage curriculum learning fine-tuning framework, by iteratively injecting knowledge from the three types of graphs into the Large Language Models. Extensive experiments show that our proposed model improves the construction of all three graph types across in-domain, OOD and counter-task data.

Backdoor attacks typically place a specific trigger on certain training data, such that the model makes prediction errors on inputs with that trigger during inference. Despite the core role of the trigger, existing studies have commonly believed a perfect match between training-inference triggers is optimal. In this paper, for the first time, we systematically explore the training-inference trigger relation, particularly focusing on their mismatch, based on a Training-Inference Trigger Intensity Manipulation (TITIM) workflow. TITIM specifically investigates the training-inference trigger intensity, such as the size or the opacity of a trigger, and reveals new insights into trigger generalization and overfitting. These new insights challenge the above common belief by demonstrating that the training-inference trigger mismatch can facilitate attacks in two practical scenarios, posing more significant security threats than previously thought. First, when the inference trigger is fixed, using training triggers with mixed intensities leads to stronger attacks than using any single intensity. For example, on CIFAR-10 with ResNet-18, mixing training triggers with 1.0 and 0.1 opacities improves the worst-case attack success rate (ASR) (over different testing opacities) of the best single-opacity attack from 10.61\% to 92.77\%. Second, intentionally using certain mismatched training-inference triggers can improve the attack stealthiness, i.e., better bypassing defenses. For example, compared to the training/inference intensity of 1.0/1.0, using 1.0/0.7 decreases the area under the curve (AUC) of the Scale-Up defense from 0.96 to 0.62, while maintaining a high attack ASR (99.65\% vs. 91.62\%). The above new insights are validated to be generalizable across different backdoor attacks, models, datasets, tasks, and (digital/physical) domains.

Low-light image enhancement aims to restore the under-exposure image captured in dark scenarios. Under such scenarios, traditional frame-based cameras may fail to capture the structure and color information due to the exposure time limitation. Event cameras are bio-inspired vision sensors that respond to pixel-wise brightness changes asynchronously. Event cameras' high dynamic range is pivotal for visual perception in extreme low-light scenarios, surpassing traditional cameras and enabling applications in challenging dark environments. In this paper, inspired by the success of the retinex theory for traditional frame-based low-light image restoration, we introduce the first methods that combine the retinex theory with event cameras and propose a novel retinex-based low-light image restoration framework named ERetinex. Among our contributions, the first is developing a new approach that leverages the high temporal resolution data from event cameras with traditional image information to estimate scene illumination accurately. This method outperforms traditional image-only techniques, especially in low-light environments, by providing more precise lighting information. Additionally, we propose an effective fusion strategy that combines the high dynamic range data from event cameras with the color information of traditional images to enhance image quality. Through this fusion, we can generate clearer and more detail-rich images, maintaining the integrity of visual information even under extreme lighting conditions. The experimental results indicate that our proposed method outperforms state-of-the-art (SOTA) methods, achieving a gain of 1.0613 dB in PSNR while reducing FLOPS by \textbf{84.28}\%.

--This is the third part of a series of studies that model the target trajectory, which describes the target state evolution over continuous time, as a sample path of a stochastic process (SP). By adopting a deterministic-stochastic decomposition framework, we decompose the learning of the trajectory SP into two sequential stages: the first fits the deterministic trend of the trajectory using a curve function of time, while the second estimates the residual stochastic component through parametric learning of either a Gaussian process (GP) or Student's-t process (StP). This leads to a Markov-free data-driven tracking approach that produces the continuous-time trajectory with minimal prior knowledge of the target dynamics. It does not only take advantage of the smooth trend of the target but also makes use of the long-term temporal correlation of both the data noise and the model fitting error . Simulations in four maneuvering target tracking scenarios have demonstrated its effectiveness and superiority in comparison with existing approaches. ARGET tracking that involves the online estimation of the trajectory of a target has been a long-standing research topic and plays a significant role in aerospace, traffic, defense, robotics, etc. [1] In essence, target tracking is more about estimating the continuous-time trajectory of the target rather than merely a finite number of point states. The continuous-time trajectory enables the acquisition of a point estimate of the state at any time in the trajectory period. However, the converse is not true. X, defined in spatio-temporal space, where X denotes the state space. Manuscript created Feb 2025; This work was supported in part by the National Natural Science Foundation of China under Grants 62422117 and 62201316 and in part by the Fundamental Research Funds for the Central Universities.