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.

Tool-integrated reasoning (TIR) enables LLM agents to solve tasks through planning, tool use, and iterative revision, but outcome-only reinforcement learning in this setting suffers from sparse, delayed rewards and weak step-level credit assignment.In long-horizon TIR trajectories, an early irrecoverable mistake can determine success or failure, making it crucial to localize the first irrecoverable step and leverage it for fine-grained credit assignment.We propose Error-Localized Policy Optimization (ELPO), which localizes the first irrecoverable step via binary-search rollout trees under a fixed rollout budget, converts the resulting tree into stable learning signals through hierarchical advantage attribution, and applies error-localized adaptive clipping to strengthen corrective updates on the critical step and its suffix.Across TIR benchmarks in math, science QA, and code execution, ELPO consistently outperforms strong Agentic RL baselines under comparable sampling budgets, with additional gains in Pass@K and Major@K scaling, rollout ranking quality, and tool-call efficiency.

Efficient lightweight neural networks are with increasing attention due to their faster reasoning speed and easier deployment on mobile devices. However, existing video pre-trained models still focus on the common ViT architecture with high latency, and few works attempt to build efficient architecture on mobile devices. This paper bridges this gap by introducing temporal structural reparameterization into an efficient image-text model and training it on a large-scale high-quality video-text dataset, resulting in an efficient video-text model that can run on mobile devices with strong zero-shot classification and retrieval capabilities, termed as MobileViCLIP. In particular, in terms of inference speed on mobile devices, our MobileViCLIP-Small is 55.4x times faster than InternVideo2-L14 and 6.7x faster than InternVideo2-S14. In terms of zero-shot retrieval performance, our MobileViCLIP-Small obtains similar performance as InternVideo2-L14 and obtains 6.9\% better than InternVideo2-S14 on MSR-VTT. The code is available at https://github.com/MCG-NJU/MobileViCLIP.

During the inference phase of Large Language Models (LLMs) with long context, a substantial portion of GPU memory is allocated to the KV cache, with memory usage increasing as the sequence length grows. To mitigate the GPU memory footprint associate with KV cache, some previous studies have discarded less important tokens based on the sparsity identified in attention scores in long context scenarios. However, we argue that attention scores cannot indicate the future importance of tokens in subsequent generation iterations, because attention scores are calculated based on current hidden states. Therefore, we propose OmniKV, a token-dropping-free and training-free inference method, which achieves a 1.68x speedup without any loss in performance. It is well-suited for offloading, significantly reducing KV cache memory usage by up to 75% with it. The core innovative insight of OmniKV is: Within a single generation iteration, there is a high degree of similarity in the important tokens identified across consecutive layers. Extensive experiments demonstrate that OmniKV achieves state-of-the-art performance across multiple benchmarks, with particularly advantages in chain-of-thoughts scenarios. OmniKV extends the maximum context length supported by a single A100 for Llama-3-8B from 128K to 450K. Our code is available at https://github.com/antgroup/OmniKV.git.

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

In this paper, we study the problem of procedure planning in instructional videos, which aims to make a plan (i.e. a sequence of actions) given the current visual observation and the desired goal. Previous works cast this as a sequence modeling problem and leverage either intermediate visual observations or language instructions as supervision to make autoregressive planning, resulting in complex learning schemes and expensive annotation costs. To avoid intermediate supervision annotation and error accumulation caused by planning autoregressively, we propose a diffusion-based framework, coined as PDPP (Projected Diffusion model for Procedure Planning), to directly model the whole action sequence distribution with task label as supervision instead. Our core idea is to treat procedure planning as a distribution fitting problem under the given observations, thus transform the planning problem to a sampling process from this distribution during inference. The diffusion-based modeling approach also effectively addresses the uncertainty issue in procedure planning. Based on PDPP, we further apply joint training to our framework to generate plans with varying horizon lengths using a single model and reduce the number of training parameters required. We instantiate our PDPP with three popular diffusion models and investigate a serious of condition-introducing methods in our framework, including condition embeddings, Mixture-of-Experts (MoEs), two-stage prediction and Classifier-Free Guidance strategy. Finally, we apply our PDPP to the Visual Planners for human Assistance (VPA) problem which requires the goal specified in natural language rather than visual observation. We conduct experiments on challenging datasets of different scales and our PDPP model achieves the state-of-the-art performance on multiple metrics, even compared with those strongly-supervised counterparts. These results further demonstrates the effectiveness and generalization ability of our model. Code and trained models are available at https://github.com/MCG-NJU/PDPP.

Current RGB-D scene recognition approaches often train two standalone backbones for RGB and depth modalities with the same Places or ImageNet pre-training. However, the pre-trained depth network is still biased by RGB-based models which may result in a suboptimal solution. In this paper, we present a single-model self-supervised hybrid pre-training framework for RGB and depth modalities, termed as CoMAE. Our CoMAE presents a curriculum learning strategy to unify the two popular self-supervised representation learning algorithms: contrastive learning and masked image modeling. Specifically, we first build a patch-level alignment task to pre-train a single encoder shared by two modalities via cross-modal contrastive learning. Then, the pre-trained contrastive encoder is passed to a multi-modal masked autoencoder to capture the finer context features from a generative perspective. In addition, our single-model design without requirement of fusion module is very flexible and robust to generalize to unimodal scenario in both training and testing phases. Extensive experiments on SUN RGB-D and NYUDv2 datasets demonstrate the effectiveness of our CoMAE for RGB and depth representation learning. In addition, our experiment results reveal that CoMAE is a data-efficient representation learner. Although we only use the small-scale and unlabeled training set for pre-training, our CoMAE pre-trained models are still competitive to the state-of-the-art methods with extra large-scale and supervised RGB dataset pre-training. Code will be released at https://github.com/MCG-NJU/CoMAE.

In this paper, we study the problem of procedure planning in instructional videos, which aims to make goal-directed plans given the current visual observations in unstructured real-life videos. Previous works cast this problem as a sequence planning problem and leverage either heavy intermediate visual observations or natural language instructions as supervision, resulting in complex learning schemes and expensive annotation costs. In contrast, we treat this problem as a distribution fitting problem. In this sense, we model the whole intermediate action sequence distribution with a diffusion model (PDPP), and thus transform the planning problem to a sampling process from this distribution. In addition, we remove the expensive intermediate supervision, and simply use task labels from instructional videos as supervision instead. Our model is a U-Net based diffusion model, which directly samples action sequences from the learned distribution with the given start and end observations. Furthermore, we apply an efficient projection method to provide accurate conditional guides for our model during the learning and sampling process. Experiments on three datasets with different scales show that our PDPP model can achieve the state-of-the-art performance on multiple metrics, even without the task supervision. Code and trained models are available at https://github.com/MCG-NJU/PDPP.

Video summarization is an essential problem in signal processing, which intends to produce a concise summary of the original video. Existing video summarization approaches regard the task as a keyframe selection problem and generally construct the frame-wise representation by combining the long-range temporal dependency with either unimodal or bimodal information. The optimal keyframe should offer the semantic summarization of the whole content by exploiting the multimodal and shot-level hierarchical natures of videos, however, such natures are not fully exploited in existing methods. In this paper, we propose to construct a more powerful and robust frame-wise representation and predict the frame-level importance score in a fair and comprehensive manner. Specifically, we propose a multimodal hierarchical shot-aware convolutional network, denoted as MHSCNet, to enhance the frame-wise representation via combining the comprehensive available multimodal information. We further design a hierarchical ShotConv network to incorporate the adaptive shot-aware frame-level representation by considering the short-range and long-range temporal dependencies. Based on the learned shot-aware representations, MHSCNet can predict the frame-level importance score in the local and global view of the video. Extensive experiments on two standard video summarization datasets demonstrate that our proposed method consistently outperforms state-of-the-arts.

Previous object detectors make predictions based on dense grid points or numerous preset anchors. Most of these detectors are trained with one-to-many label assignment strategies. On the contrary, recent query-based object detectors are based a sparse set of learnable queries refined by a series of decoder layers. The one-to-one label assignment is independently applied on each layer for deep supervision during training. Despite the great success of query-based object detection, however, this vanilla one-to-one label assignment strategy requires the detectors to have strong fine-grained discrimination and modeling capacity. In this paper, we propose a new query-based object detector with cross-stage interaction, coined as StageInter-actor. During the forward pass, we come up with an efficient way to improve this modeling ability by reusing dynamic operators with lightweight adapters. As for the label assignment, a cross-stage label assigner is designed to improve the one-to-one label assignment. With this assigner, the training target class labels are gathered across stages and then reallocated to proper predictions at each decoder layer. On MS COCO benchmark, our model improves the baseline counterpart by 2.2 AP, and achieves a 44.8 AP with ResNet-50 as backbone, 100 queries and 12 training epochs. With longer training time and 300 queries, StageIn-teractor achieves 51.3 AP and 52.7 AP with ResNeXt-101-DCN and Swin-S, respectively. The code and models are made available at https://github.com/MCG-NJU/StageInteractor.

Sequential recommendation (SR) learns users’ preferences by capturing the sequential patterns from users’ behaviors evolution. As discussed in many works, user–item interactions of SR generally present the intrinsic power-law distribution, which can be ascended to hierarchy-like structures. Previous methods usually handle such hierarchical information by making user–item sectionalization empirically under Euclidean space, which may cause distortion of user–item representation in real online scenarios. In this article, we propose a Poincaré-based heterogeneous graph neural network named Poincaré Heterogeneous Graph Neural Networks for Sequential Recommendation (PHGR) to model the sequential pattern information as well as hierarchical information contained in the data of SR scenarios simultaneously. Specifically, for the purpose of explicitly capturing the hierarchical information, we first construct a weighted user–item heterogeneous graph by aliening all the user–item interactions to improve the perception domain of each user from a global view. Then the output of the global representation would be used to complement the local directed item–item homogeneous graph convolution. By defining a novel hyperbolic inner product operator, the global and local graph representation learning are directly conducted in Poincaré ball instead of commonly used projection operation between Poincaré ball and Euclidean space, which could alleviate the cumulative error issue of general bidirectional translation process. Moreover, for the purpose of explicitly capturing the sequential dependency information, we design two types of temporal attention operations under Poincaré ball space. Empirical evaluations on datasets from the public and financial industry show that PHGR outperforms several comparison methods.

This work aims at improving instance retrieval with self-supervision. We find that fine-tuning using the recently developed self-supervised learning (SSL) methods, such as SimCLR and MoCo, fails to improve the performance of instance retrieval. In this work, we identify that the learnt representations for instance retrieval should be invariant to large variations in viewpoint and background etc., whereas self-augmented positives applied by the current SSL methods can not provide strong enough signals for learning robust instance-level representations. To overcome this problem, we propose InsCLR, a new SSL method that builds on the instance-level contrast, to learn the intra-class invariance by dynamically mining meaningful pseudo positive samples from both mini-batches and a memory bank during training. Extensive experiments demonstrate that InsCLR achieves similar or even better performance than the state-of-the-art SSL methods on instance retrieval. Code is available at https://github.com/zeludeng/insclr.

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.

Traditional object detectors employ the dense paradigm of scanning over locations and scales in an image. The recent query-based object detectors break this convention by decoding image features with a set of learnable queries. However, this paradigm still suffers from slow convergence, limited performance, and design complexity of extra networks between backbone and decoder. In this paper, we find that the key to these issues is the adaptability of decoders for casting queries to varying objects. Accordingly, we propose a fast-converging query-based detector, named AdaMixer, by improving the adaptability of query-based decoding processes in two aspects. First, each query adaptively samples features over space and scales based on estimated offsets, which allows AdaMixer to efficiently attend to the coherent regions of objects. Then, we dynamically decode these sampled features with an adaptive MLP-Mixer under the guidance of each query. Thanks to these two critical designs, AdaMixer enjoys architectural simplicity without requiring dense attentional encoders or explicit pyramid networks. On the challenging MS COCO benchmark, AdaMixer with ResNet-50 as the backbone, with 12 training epochs, reaches up to 45.0 AP on the validation set along with 27.9 APs in detecting small objects. With the longer training scheme, AdaMixer with ResNeXt-101-DCN and Swin-S reaches 49.5 and 51.3 AP. Our work sheds light on a simple, accurate, and fast converging architecture for query-based object detectors. The code is made available at https://github.com/MCG-NJU/AdaMixer.

Though widely used in image classification, convolutional neural networks (CNNs) are prone to noise interruptions, i.e. the CNN output can be drastically changed by small image noise. To improve the noise robustness, we try to integrate CNNs with wavelet by replacing the common down-sampling (max-pooling, strided-convolution, and average pooling) with discrete wavelet transform (DWT). We firstly propose general DWT and inverse DWT (IDWT) layers applicable to various orthogonal and biorthogonal discrete wavelets like Haar, Daubechies, and Cohen, etc., and then design wavelet integrated CNNs (WaveCNets) by integrating DWT into the commonly used CNNs (VGG, ResNets, and DenseNet). During the down-sampling, WaveCNets apply DWT to decompose the feature maps into the low-frequency and high-frequency components. Containing the main information including the basic object structures, the low-frequency component is transmitted into the following layers to generate robust high-level features. The high-frequency components are dropped to remove most of the data noises. The experimental results show that WaveCNets achieve higher accuracy on ImageNet than various vanilla CNNs. We have also tested the performance of WaveCNets on the noisy version of ImageNet, ImageNet-C and six adversarial attacks, the results suggest that the proposed DWT/IDWT layers could provide better noise-robustness and adversarial robustness. When applying WaveCNets as backbones, the performance of object detectors (i.e., faster R-CNN and RetinaNet) on COCO detection dataset are consistently improved. We believe that suppression of aliasing effect, i.e. separation of low frequency and high frequency information, is the main advantages of our approach. The code of our DWT/IDWT layer and different WaveCNets are available at https://github.com/CVI-SZU/WaveCNet.

In this work, we propose Knowledge Integration Networks (referred as KINet) for video action recognition. KINet is capable of aggregating meaningful context features which are of great importance to identifying an action, such as human information and scene context. We design a three-branch architecture consisting of a main branch for action recognition, and two auxiliary branches for human parsing and scene recognition which allow the model to encode the knowledge of human and scene for action recognition. We explore two pre-trained models as teacher networks to distill the knowledge of human and scene for training the auxiliary tasks of KINet. Furthermore, we propose a two-level knowledge encoding mechanism which contains a Cross Branch Integration (CBI) module for encoding the auxiliary knowledge into medium-level convolutional features, and an Action Knowledge Graph (AKG) for effectively fusing high-level context information. This results in an end-to-end trainable framework where the three tasks can be trained collaboratively, allowing the model to compute strong context knowledge efficiently. The proposed KINet achieves the state-of-the-art performance on a large-scale action recognition benchmark Kinetics-400, with a top-1 accuracy of 77.8%. We further demonstrate that our KINet has strong capability by transferring the Kinetics-trained model to UCF-101, where it obtains 97.8% top-1 accuracy.

Convolutional Neural Networks (CNNs) are generally prone to noise interruptions, i.e., small image noise can cause drastic changes in the output. To suppress the noise effect to the final predication, we enhance CNNs by replacing max-pooling, strided-convolution, and average-pooling with Discrete Wavelet Transform (DWT). We present general DWT and Inverse DWT (IDWT) layers applicable to various wavelets like Haar, Daubechies, and Cohen, etc., and design wavelet integrated CNNs (WaveCNets) using these layers for image classification. In WaveCNets, feature maps are decomposed into the low-frequency and high-frequency components during the down-sampling. The low-frequency component stores main information including the basic object structures, which is transmitted into the subsequent layers to extract robust high-level features. The high-frequency components, containing most of the data noise, are dropped during inference to improve the noise-robustness of the WaveCNets. Our experimental results on ImageNet and ImageNet-C (the noisy version of ImageNet) show that WaveCNets, the wavelet integrated versions of VGG, ResNets, and DenseNet, achieve higher accuracy and better noise-robustness than their vanilla versions.

Multi-label image classification has achieved remarkable progress thanks to deep convolutional neural networks (CNNs). In this paper, we propose a Decouple Network (DecoupleNet) which is an end-to-end CNN-based framework able to trade off class-level feature independence and relevance during training. The proposed DecoupleNet is able to decouple category-wise independence and relevance with image-level supervision. We design a category-wise space-to-depth module with a spatial pooling strategy to exploit more meaningful convolutional features. They are integrated with class-wise correlated information which is automatically learned via a new self-attention mechanism. We conduct extensive experiments on two large-scale benchmarks: the MS-COCO and the NUS-WIDE, where the proposed DecoupleNet obtains impressive performance compared favorably against the state-of-the-art methods on multi-label image classification.

Weakly-supervised instance segmentation aims to detect and segment object instances precisely, given image-level labels only. Unlike previous methods which are composed of multiple offline stages, we propose Sequential Label Propagation and Enhancement Networks (referred as Label-PEnet) that progressively transforms image-level labels to pixel-wise labels in a coarse-to-fine manner. We design four cascaded modules including multi-label classification, object detection, instance refinement and instance segmentation, which are implemented sequentially by sharing the same backbone. The cascaded pipeline is trained alternatively with a curriculum learning strategy that generalizes labels from high level images to low-level pixels gradually with increasing accuracy. In addition, we design a proposal calibration module to explore the ability of classification networks to find key pixels that identify object parts, which serves as a post validation strategy running in the inverse order. We evaluate the efficiency of our Label-PEnet in mining instance masks on standard benchmarks: PASCAL VOC 2007 and 2012. Experimental results show that Label-PEnet outperforms the state-of-art algorithms by a clear margin, and obtains comparable performance even with fully supervised approaches.

Convolutional neural networks (CNNs) have recently achieved remarkable successes in various image classification and understanding tasks. The deep features obtained at the top fully connected layer of the CNN (FC-features) exhibit rich global semantic information and are extremely effective in image classification. On the other hand, the convolutional features in the middle layers of the CNN also contain meaningful local information, but are not fully explored for image representation. In this paper, we propose a novel locally supervised deep hybrid model (LS-DHM) that effectively enhances and explores the convolutional features for scene recognition. First, we notice that the convolutional features capture local objects and fine structures of scene images, which yield important cues for discriminating ambiguous scenes, whereas these features are significantly eliminated in the highly compressed FC representation. Second, we propose a new local convolutional supervision layer to enhance the local structure of the image by directly propagating the label information to the convolutional layers. Third, we propose an efficient Fisher convolutional vector (FCV) that successfully rescues the orderless mid-level semantic information (e.g., objects and textures) of scene image. The FCV encodes the large-sized convolutional maps into a fixed-length mid-level representation, and is demonstrated to be strongly complementary to the high-level FC-features. Finally, both the FCV and FC-features are collaboratively employed in the LS-DHM representation, which achieves outstanding performance in our experiments. It obtains 83.75% and 67.56% accuracies, respectively, on the heavily benchmarked MIT Indoor67 and SUN397 data sets, advancing the state-of-the-art substantially.

Convolutional neural networks (CNNs) have made remarkable progress on scene recognition, partially due to these recent large-scale scene datasets, such as the Places and Places2. Scene categories are often defined by multi-level information, including local objects, global layout, and background environment, thus leading to large intra-class variations. In addition, with the increasing number of scene categories, label ambiguity has become another crucial issue in large-scale classification. This paper focuses on large-scale scene recognition and makes two major contributions to tackle these issues. First, we propose a multi-resolution CNN architecture that captures visual content and structure at multiple levels. The multi-resolution CNNs are composed of coarse resolution CNNs and fine resolution CNNs, which are complementary to each other. Second, we design two knowledge guided disambiguation techniques to deal with the problem of label ambiguity: 1) we exploit the knowledge from the confusion matrix computed on validation data to merge ambiguous classes into a super category and 2) we utilize the knowledge of extra networks to produce a soft label for each image. Then, the super categories or soft labels are employed to guide CNN training on the Places2. We conduct extensive experiments on three large-scale image datasets (ImageNet, Places, and Places2), demonstrating the effectiveness of our approach. Furthermore, our method takes part in two major scene recognition challenges, and achieves the second place at the Places2 challenge in ILSVRC 2015, and the first place at the LSUN challenge in CVPR 2016. Finally, we directly test the learned representations on other scene benchmarks, and obtain the new state-of-the-art results on the MIT Indoor67 (86.7%) and SUN397 (72.0%). We release the code and models at https://github.com/wanglimin/MRCNN-Scene-Recognition.