Diffusion Transformers (DiTs) have achieved state-of-the-art performance in generative modeling, yet their high computational cost hinders real-time deployment. While feature caching offers a promising training-free acceleration solution by exploiting temporal redundancy, existing methods suffer from two key limitations: (1) uniform caching intervals fail to align with the non-uniform temporal dynamics of DiT, and (2) naive feature reuse with excessively large caching intervals can lead to severe error accumulation. In this work, we analyze the evolution of DiT features during denoising and reveal that both feature changes and error propagation are highly time- and depth-varying. Motivated by this, we propose ProCache, a training-free dynamic feature caching framework that addresses these issues via two core components: (i) a constraint-aware caching pattern search module that generates non-uniform activation schedules through offline constrained sampling, tailored to the model’s temporal characteristics; and (ii) a selective computation module that selectively compute within deep blocks and high-importance tokens for cached segments to mitigate error accumulation with minimal overhead. Extensive experiments on PixArt-alpha and DiT demonstrate that ProCache achieves up to 1.96 times and 2.90 times acceleration with negligible quality degradation, significantly outperforming prior caching-based methods.

Partially Relevant Video Retrieval (PRVR) aims to retrieve untrimmed videos containing relevant moments for a given text query. This task is extremely challenging, as untrimmed videos often include numerous actions and objects unrelated to the query. However, existing methods usually struggle with fine-grained action-object modeling, limiting their retrieval performance. To tackle this challenge, we introduce Action-and-object Aware Alignment for Partially Relevant Video Retrieval (A3PRVR), a dual-branch framework designed to enhance retrieval by improving the modeling of action-object relationships. Specifically, we propose a Query-specific Deformable Temporal Attention (Q-DTA) module to effectively capture action-relevant object information in video features, while filtering out irrelevant content. Additionally, we propose an action-and-object aware alignment module to enable fine-grained textual understanding and video-text alignment. It uses action- and object-aware contrastive losses to enhance the model's sensitivity to action-object distinctions in the text query. Compared to state-of-the-art methods, A3PRVR achieves an average relative gain of 6.5% in SumR across the Charades-STA, ActivityNet-Caption, and TVR datasets.

Zeng You, Yaofo Chen, Qiuwu Chen, Ying Sun, Shuhai Zhang, Yingjian Li, Yaowei Wang, Mingkui Tan. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Transformer-based Large Language Models (LLMs) have exhibited remarkable success in extensive tasks primarily attributed to self-attention mechanism, which requires a token to consider all preceding tokens as its context to compute attention. However, when the context length L becomes very large (e.g., 128K), the amount of potentially redundant information in the context tends to increase. The redundant context not only hampers the modeling representation performance but also incurs unnecessary computational and storage overhead. In this paper, we propose a plug-and-play Core Context Aware (CCA) Attention for efficient long-context modeling, comprising two complementary modules: 1) Globality-aware pooling module groups input tokens and dynamically compresses each group into one core token based on their significance. In this way, our method automatically focuses and strengthens core context while diminishing redundancy during the learning process, leading to effective long-term dependency modeling. 2) Locality-preserving module incorporates neighboring tokens to preserve local context for detailed representation. Notably, our CCA-Attention is able to replace the self-attention module in existing LLMs with minimal fine-tuning cost. Extensive experimental results show the superiority of our method in both long-context modeling and computational efficiency over state-of-the-art methods.

Transformer-based large language models (LLMs) excel in natural language processing tasks by capturing long-range dependencies through self-attention mechanisms. However, long-context modeling faces significant computational inefficiencies due to redundant attention computations: while attention weights are often sparse, all tokens consume equal computational resources. In this paper, we reformulate traditional probabilistic sequence modeling as a supervised learning task, enabling the separation of relevant and irrelevant tokens and providing a clearer understanding of redundancy. Based on this reformulation, we theoretically analyze attention sparsity, revealing that only a few tokens significantly contribute to predictions. Building on this, we formulate attention optimization as a linear coding problem and propose a group coding strategy, theoretically showing its ability to improve robustness against random noise and enhance learning efficiency. Motivated by this, we propose Dynamic Group Attention (DGA), which leverages the group coding to explicitly reduce redundancy by aggregating less important tokens during attention computation. Empirical results show that our DGA significantly reduces computational costs while maintaining competitive performance.

Long video understanding has become a critical task in computer vision, driving advancements across numerous applications from surveillance to content retrieval. Existing video understanding methods suffer from two challenges when dealing with long video understanding: intricate long-context relationship modeling and interference from redundancy. To tackle these challenges, we introduce Fine-Detailed Video Story generation (FDVS), which interprets long videos into detailed textual representations. Specifically, to achieve fine-grained modeling of long-temporal content, we propose a Bottom-up Video Interpretation Mechanism that progressively interprets video content from clips to video. To avoid interference from redundant information in videos, we introduce a Semantic Redundancy Reduction mechanism that removes redundancy at both the visual and textual levels. Our method transforms long videos into hierarchical textual representations that contain multi-granularity information of the video. With these representations, FDVS is applicable to various tasks without any fine-tuning. We evaluate the proposed method across eight datasets spanning three tasks. The performance demonstrates the effectiveness and versatility of our method.

Zero-shot skeleton-based action recognition aims to classify unseen skeleton-based human actions without prior exposure to such categories during training. This task is extremely challenging due to the difficulty in generalizing from known to unknown actions. Previous studies typically use two-stage training: pre-training skeleton encoders on seen action categories using cross-entropy loss and then aligning pre-extracted skeleton and text features, enabling knowledge transfer to unseen classes through skeleton-text alignment and language models' generalization. However, their efficacy is hindered by 1) insufficient discrimination for skeleton features, as the fixed skeleton encoder fails to capture necessary alignment information for effective skeleton-text alignment; 2) the neglect of alignment bias between skeleton and unseen text features during testing. To this end, we propose a prototype-guided feature alignment paradigm for zero-shot skeleton-based action recognition, termed PGFA. Specifically, we develop an end-to-end cross-modal contrastive training framework to improve skeleton-text alignment, ensuring sufficient discrimination for skeleton features. Additionally, we introduce a prototype-guided text feature alignment strategy to mitigate the adverse impact of the distribution discrepancy during testing. We provide a theoretical analysis to support our prototype-guided text feature alignment strategy and empirically evaluate our overall PGFA on three well-known datasets. Compared with the top competitor SMIE method, our PGFA achieves absolute accuracy improvements of 22.96%, 12.53%, and 18.54% on the NTU-60, NTU-120, and PKU-MMD datasets, respectively.

Videofluoroscopic swallowing study (VFSS) visualizes the swallowing movement by using X-ray fluoroscopy, which is the most widely used method for dysphagia examination. To better facilitate swallowing assessment, the temporal parameter is one of the most important indicators. However, most information of that acquire is hand-crafted and elaborated, which is time-consuming and difficult to ensure objectivity and accuracy. In this article, we propose to formulate this task as a temporal action localization task and solve it using deep neural networks. However, the action of VFSS has the following characteristics such as small motion targets, small action amplitudes, large sample variances, short duration, and variations in duration. Furthermore, all existing methods often rely on daily behaviors, which makes locating and recognizing micro-actions more challenging. To address the above issues, we first collect and annotate the VFSS micro-action dataset, which includes 847 VFSS data from 71 subjects, due to the lack of benchmarks. We then introduce a coarse-to-fine mechanism to handle the short and repeated nature of micro-actions, which can significantly enhancing micro-action localization accuracy. Moreover, we propose a Variable-Size Window Generator method, which improves the model's characterization performance and addresses the issue of different action timings, leading to further improvements in localization accuracy. The results of our experiments demonstrate the superiority of our method, with significantly improved performance (46.10% vs. 37.70%).

Alzheimer's disease(AD) is a neurodegenerative disease that progresses slowly but worsens gradually, also, the most common kinds of dementia. Clinically, the diagnosis of AD is mainly based on rating scales and neuroimaging technology which is invasive, costly and time-consuming. Other than that, the clinical pathology has become irreversible when neuroimaging characteristics appear. It is imperative to develop new noninvasive methods for early diagnosis of AD. Several studies indicated the probable association of cognitive decline with gait changes might shed light on potential features for distinction of AD. This paper aims to exploit the feasibility of gait features for early diagnosis of mild cognitive impairment(MCI) and AD by using machine learning methods. A device-free AD detection system is built, with a natural undisturbed gait collecting system and a well-performed Long Short-Term Memory(LSTM) based model, in this article. Moreover, it can serve as a simplified, non-invasive, and highly accurate clinical auxiliary tool for early diagnosis and distinction of AD. Experimental results showed a 90.48%, 92.00%, and 88.24% in accuracy, sensitivity, and specificity respectively for distinguishing AD by using the method with LSTM based model. Furthermore, the gait cycle and stride length in MCI or AD were more variable than in healthy controls through redefining and calculating the gait features with skeleton data obtained by Kinect devices.