Temporal graphs are essential for modeling complex real-world systems, such as social interactions, financial transactions, and recommendation systems, but the high computational cost and model complexity of dynamic graph neural networks (DGNNs) pose significant challenges for practical deployment. Although various pruning and sampling techniques have proven effective in accelerating static GNNs, they fall short in dynamic settings due to temporal dependencies in evolving graph structures. To address these challenges, we propose TrimDG, a general framework that accelerates DGNNs by eliminating both static and runtime redundancies. For static redundancy, we introduce a novel node influence metric, Temporal Personalized PageRank (TPP), to prune less informative nodes, and employ temporal binning to remove redundant events. For runtime redundancy during training, we develop an adaptive sampling strategy guided by graph information bottleneck and further reduce sampling frequency through temporal batch selector and sampling cache. Theoretical analysis supports our design, and experiments on real-world datasets show that TrimDG reduces runtime by an average of 83.49% across diverse DGNN backbones, while maintaining strong predictive performance, demonstrating both its efficiency and generalizability.

Public Speaking Anxiety Detection (PSAD) is a complex and challenging task that involves detecting anxiety through diverse multimodal cues. While deep neural networks have achieved remarkable success in this task, their performance tends to degrade significantly under distribution shifts, especially in Out-Of-Distribution (OOD) scenarios. Given the substantial impact of latent contexts in PSAD, it is crucial to explore how contexts influence OOD generalization. Causality plays a fundamental role in uncovering the relationships between multimodal data and labels. Understanding these causal relationships helps tackle the OOD generalization challenge via causal invariance under context-driven shifts. In this work, we propose a novel Context-Aware Multimodal Adversarial Learning (CAMAL) framework to model and debias complex multimodal contexts in PSAD. A dual adversarial learning strategy jointly optimizes the causal independence between multimodal causal features and contexts. Experimental results on two benchmark datasets demonstrate the superiority of our approach over state-of-the-art baselines under distribution shifts.

Recent advancements in document understanding have been dominated by leveraging large language models (LLMs) and multimodal large models. However, enabling LLMs to comprehend complex document layouts and structural information often necessitates intricate network modifications or costly pre-training, limiting their practical applicability. In this paper, we introduce Group Position Embedding (GPE), a novel and efficient technique to enhance the layout understanding capabilities of LLMs without architectural changes or additional pre-training. GPE achieves this by strategically grouping the attention heads and feeding each group with distinct positional embeddings, effectively encoding layout information relevant to document comprehension. This simple yet powerful method allows for effective integration of layout information within the existing LLM framework. We evaluate GPE against several competitive baselines across five mainstream document tasks. We also introduce a challenging benchmark called BLADE, specifically designed to assess layout comprehension. Extensive experiments on both established and BLADE benchmarks confirm the efficacy of GPE in significantly advancing the state-of-the-art in document understanding. Our code is available at https://github.com/antgroup/GroupPositionEmbedding.git

Assessment of speech presentation skills via multimodal learning has attracted significant attention. However, existing assessment methods in this area often focus on single-dimensional criteria, overlooking the complex interplay among multiple factors that shape a speaker’s performance. Cross-domain transfer approaches also struggle to effectively model the dependencies between these key dimensions. To address these limitations, we propose the cGAN-Prompt Guided Multi-Task Decoupling (cG-PMTD) model, which employs a prompt-driven mechanism to enhance positive information transfer and mitigate interference across multiple evaluation dimensions. Additionally, we introduce the Classroom Public Speaking Assessment (CPSA) dataset, a large-scale, fine-grained multimodal resource designed specifically for multidimensional evaluation tasks. Experimental results on CPSA demonstrate that cG-PMTD significantly outperforms state-of-the-art baselines, offering a more comprehensive and accurate assessment of public speaking abilities. The code and dataset publicly available at https://github.com/ZHxiong-sc/cG-PMTD-and-CSPA-dataset.

In this paper, we present a new cross-architecture contrastive learning (CACL) framework for self-supervised video representation learning. CACL consists of a 3D CNN and a video transformer which are used in parallel to generate diverse positive pairs for contrastive learning. This allows the model to learn strong representations from such diverse yet meaningful pairs. Furthermore, we introduce a temporal self-supervised learning module able to predict an Edit distance explicitly between two video sequences in the temporal order. This enables the model to learn a rich temporal representation that compensates strongly to the video-level representation learned by the CACL. We evaluate our method on the tasks of video retrieval and action recognition on UCF101 and HMDB51 datasets, where our method achieves excellent performance, surpassing the state-of-the-art methods such as VideoMoCo [23] and MoCo+BE [34] by a large margin.