Jun Yu

Ce Zhou, Qian Li, Chen Li et al. · allen-ai

Pretrained Foundation Models (PFMs) are regarded as the foundation for various downstream tasks with different data modalities. A PFM (e.g., BERT, ChatGPT, and GPT-4) is trained on large-scale data which provides a reasonable parameter initialization for a wide range of downstream applications. BERT learns bidirectional encoder representations from Transformers, which are trained on large datasets as contextual language models. Similarly, the generative pretrained transformer (GPT) method employs Transformers as the feature extractor and is trained using an autoregressive paradigm on large datasets. Recently, ChatGPT shows promising success on large language models, which applies an autoregressive language model with zero shot or few shot prompting. The remarkable achievements of PFM have brought significant breakthroughs to various fields of AI. Numerous studies have proposed different methods, raising the demand for an updated survey. This study provides a comprehensive review of recent research advancements, challenges, and opportunities for PFMs in text, image, graph, as well as other data modalities. The review covers the basic components and existing pretraining methods used in natural language processing, computer vision, and graph learning. Additionally, it explores advanced PFMs used for different data modalities and unified PFMs that consider data quality and quantity. The review also discusses research related to the fundamentals of PFMs, such as model efficiency and compression, security, and privacy. Finally, the study provides key implications, future research directions, challenges, and open problems in the field of PFMs. Overall, this survey aims to shed light on the research of the PFMs on scalability, security, logical reasoning ability, cross-domain learning ability, and the user-friendly interactive ability for artificial general intelligence.

Zhou Yu, Xuecheng Ouyang, Zhenwei Shao et al.

Knowledge-based visual question answering (VQA) requires external knowledge beyond the image to answer the question. Early studies retrieve required knowledge from explicit knowledge bases (KBs), which often introduces irrelevant information to the question, hence restricting the performance of their models. Recent works have resorted to using a powerful large language model (LLM) as an implicit knowledge engine to acquire the necessary knowledge for answering. Despite the encouraging results achieved by these methods, we argue that they have not fully activated the capacity of the \emph{blind} LLM as the provided textual input is insufficient to depict the required visual information to answer the question. In this paper, we present Prophet -- a conceptually simple, flexible, and general framework designed to prompt LLM with answer heuristics for knowledge-based VQA. Specifically, we first train a vanilla VQA model on a specific knowledge-based VQA dataset without external knowledge. After that, we extract two types of complementary answer heuristics from the VQA model: answer candidates and answer-aware examples. The two types of answer heuristics are jointly encoded into a formatted prompt to facilitate the LLM's understanding of both the image and question, thus generating a more accurate answer. By incorporating the state-of-the-art LLM GPT-3, Prophet significantly outperforms existing state-of-the-art methods on four challenging knowledge-based VQA datasets. Prophet is general that can be instantiated with the combinations of different VQA models (i.e., both discriminative and generative ones) and different LLMs (i.e., both commercial and open-source ones). Moreover, Prophet can also be integrated with modern large multimodal models in different stages, which is named Prophet++, to further improve the capabilities on knowledge-based VQA tasks.

Jun Yu, Zhaoming Kong, Liang Zhan et al.

The assessment of Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI) associated with brain changes remains a challenging task. Recent studies have demonstrated that combination of multi-modality imaging techniques can better reflect pathological characteristics and contribute to more accurate diagnosis of AD and MCI. In this paper, we propose a novel tensor-based multi-modality feature selection and regression method for diagnosis and biomarker identification of AD and MCI from normal controls. Specifically, we leverage the tensor structure to exploit high-level correlation information inherent in the multi-modality data, and investigate tensor-level sparsity in the multilinear regression model. We present the practical advantages of our method for the analysis of ADNI data using three imaging modalities (VBM- MRI, FDG-PET and AV45-PET) with clinical parameters of disease severity and cognitive scores. The experimental results demonstrate the superior performance of our proposed method against the state-of-the-art for the disease diagnosis and the identification of disease-specific regions and modality-related differences. The code for this work is publicly available at https://github.com/junfish/BIOS22.

Zhaoming Kong, Fangxi Deng, Haomin Zhuang et al.

The advancement of imaging devices and countless images generated everyday pose an increasingly high demand on image denoising, which still remains a challenging task in terms of both effectiveness and efficiency. To improve denoising quality, numerous denoising techniques and approaches have been proposed in the past decades, including different transforms, regularization terms, algebraic representations and especially advanced deep neural network (DNN) architectures. Despite their sophistication, many methods may fail to achieve desirable results for simultaneous noise removal and fine detail preservation. In this paper, to investigate the applicability of existing denoising techniques, we compare a variety of denoising methods on both synthetic and real-world datasets for different applications. We also introduce a new dataset for benchmarking, and the evaluations are performed from four different perspectives including quantitative metrics, visual effects, human ratings and computational cost. Our experiments demonstrate: (i) the effectiveness and efficiency of representative traditional denoisers for various denoising tasks, (ii) a simple matrix-based algorithm may be able to produce similar results compared with its tensor counterparts, and (iii) the notable achievements of DNN models, which exhibit impressive generalization ability and show state-of-the-art performance on various datasets. In spite of the progress in recent years, we discuss shortcomings and possible extensions of existing techniques. Datasets, code and results are made publicly available and will be continuously updated at https://github.com/ZhaomingKong/Denoising-Comparison.

Yijie Lin, Mouxing Yang, Jun Yu et al.

In this paper, we study a novel and widely existing problem in graph matching (GM), namely, Bi-level Noisy Correspondence (BNC), which refers to node-level noisy correspondence (NNC) and edge-level noisy correspondence (ENC). In brief, on the one hand, due to the poor recognizability and viewpoint differences between images, it is inevitable to inaccurately annotate some keypoints with offset and confusion, leading to the mismatch between two associated nodes, i.e., NNC. On the other hand, the noisy node-to-node correspondence will further contaminate the edge-to-edge correspondence, thus leading to ENC. For the BNC challenge, we propose a novel method termed Contrastive Matching with Momentum Distillation. Specifically, the proposed method is with a robust quadratic contrastive loss which enjoys the following merits: i) better exploring the node-to-node and edge-to-edge correlations through a GM customized quadratic contrastive learning paradigm; ii) adaptively penalizing the noisy assignments based on the confidence estimated by the momentum teacher. Extensive experiments on three real-world datasets show the robustness of our model compared with 12 competitive baselines. The code is available at https://github.com/XLearning-SCU/2023-ICCV-COMMON.

Zi Wang, Katsuya Hotta, Yawen Zou et al.

Few-shot cross-category 3D anomaly detection aims to determine whether an unknown point cloud belongs to a target normal category using only a few normal references. Existing training-based methods usually require category-wise optimization, while recent training-free methods based on multi-view CLIP visual features mainly rely on visual similarity and may be confused by geometrically similar categories. In this paper, we propose VT-3DAD, a training-free framework for cross-category 3D anomaly detection via Visual-Text Normal Space Alignment. Given few-shot normal references and a test point cloud, VT-3DAD first generates realistic multi-view depth maps and extracts view-wise features using a frozen CLIP visual encoder. The visual branch measures reference-test deviation in the multi-view feature space. In parallel, depth-aware and 3D-aware prompts are encoded by the frozen CLIP text encoder to construct textual normal anchors, which provide semantic normality constraints for the target category. The final anomaly score is obtained by fusing visual deviation from normal references and semantic deviation from the textual normal space. Experiments on the ShapeNetPart dataset demonstrate that VT-3DAD achieves state-of-the-art performance. In particular, VT-3DAD improves the one-shot average AUC-ROC from 92.49% to 94.80% compared with the visual-only baseline, while also reducing the average standard deviation from 5.64 to 3.41.

Chaojian Yu, Bo Han, Li Shen et al.

Robust overfitting widely exists in adversarial training of deep networks. The exact underlying reasons for this are still not completely understood. Here, we explore the causes of robust overfitting by comparing the data distribution of \emph{non-overfit} (weak adversary) and \emph{overfitted} (strong adversary) adversarial training, and observe that the distribution of the adversarial data generated by weak adversary mainly contain small-loss data. However, the adversarial data generated by strong adversary is more diversely distributed on the large-loss data and the small-loss data. Given these observations, we further designed data ablation adversarial training and identify that some small-loss data which are not worthy of the adversary strength cause robust overfitting in the strong adversary mode. To relieve this issue, we propose \emph{minimum loss constrained adversarial training} (MLCAT): in a minibatch, we learn large-loss data as usual, and adopt additional measures to increase the loss of the small-loss data. Technically, MLCAT hinders data fitting when they become easy to learn to prevent robust overfitting; philosophically, MLCAT reflects the spirit of turning waste into treasure and making the best use of each adversarial data; algorithmically, we designed two realizations of MLCAT, and extensive experiments demonstrate that MLCAT can eliminate robust overfitting and further boost adversarial robustness.

Zhuo Huang, Miaoxi Zhu, Xiaobo Xia et al.

Robust generalization aims to tackle the most challenging data distributions which are rare in the training set and contain severe noises, i.e., photon-limited corruptions. Common solutions such as distributionally robust optimization (DRO) focus on the worst-case empirical risk to ensure low training error on the uncommon noisy distributions. However, due to the over-parameterized model being optimized on scarce worst-case data, DRO fails to produce a smooth loss landscape, thus struggling on generalizing well to the test set. Therefore, instead of focusing on the worst-case risk minimization, we propose SharpDRO by penalizing the sharpness of the worst-case distribution, which measures the loss changes around the neighbor of learning parameters. Through worst-case sharpness minimization, the proposed method successfully produces a flat loss curve on the corrupted distributions, thus achieving robust generalization. Moreover, by considering whether the distribution annotation is available, we apply SharpDRO to two problem settings and design a worst-case selection process for robust generalization. Theoretically, we show that SharpDRO has a great convergence guarantee. Experimentally, we simulate photon-limited corruptions using CIFAR10/100 and ImageNet30 datasets and show that SharpDRO exhibits a strong generalization ability against severe corruptions and exceeds well-known baseline methods with large performance gains.

Zhengyao Fang, Pengyuan Lyu, Chengquan Zhang et al.

Vision-language models (VLMs) face significant computational inefficiencies caused by excessive generation of visual tokens. While prior work shows that a large fraction of visual tokens are redundant, existing compression methods struggle to balance importance preservation and information diversity. To address this, we propose PruneSID, a training-free Synergistic Importance-Diversity approach featuring a two-stage pipeline: (1) Principal Semantic Components Analysis (PSCA) for clustering tokens into semantically coherent groups, ensuring comprehensive concept coverage, and (2) Intra-group Non-Maximum Suppression (NMS) for pruning redundant tokens while preserving key representative tokens within each group. Additionally, PruneSID incorporates an information-aware dynamic compression ratio mechanism that optimizes token compression rates based on image complexity, enabling more effective average information preservation across diverse scenes. Extensive experiments demonstrate state-of-the-art performance, achieving 96.3% accuracy on LLaVA-1.5 with only 11.1% token retention, and 92.8% accuracy at extreme compression rates (5.6%) on LLaVA-NeXT, outperforming prior methods by 2.5% with 7.8 $\times$ faster prefilling speed compared to the original model. Our framework generalizes across diverse VLMs and both image and video modalities, showcasing strong cross-modal versatility. Code is available at https://github.com/ZhengyaoFang/PruneSID.

Zhenying Fang, Jun Yu, Richang Hong

Temporal action detection aims to recognize the action category and determine each action instance's starting and ending time in untrimmed videos. The mixed methods have achieved remarkable performance by seamlessly merging anchor-based and anchor-free approaches. Nonetheless, there are still two crucial issues within the mixed framework: (1) Brute-force merging and handcrafted anchor design hinder the substantial potential and practicality of the mixed methods. (2) Within-category predictions show a significant abundance of false positives. In this paper, we propose a novel Boundary Discretization and Reliable Classification Network (BDRC-Net) that addresses the issues above by introducing boundary discretization and reliable classification modules. Specifically, the boundary discretization module (BDM) elegantly merges anchor-based and anchor-free approaches in the form of boundary discretization, eliminating the need for the traditional handcrafted anchor design. Furthermore, the reliable classification module (RCM) predicts reliable global action categories to reduce false positives. Extensive experiments conducted on different benchmarks demonstrate that our proposed method achieves competitive detection performance. The code will be released at https://github.com/zhenyingfang/BDRC-Net.

Jun Yu, Hao Chang, Keda Lu et al.

Multi-modal aerial view object classification (MAVOC) in Automatic target recognition (ATR), although an important and challenging problem, has been under studied. This paper firstly finds that fine-grained data, class imbalance and various shooting conditions preclude the representational ability of general image classification. Moreover, the MAVOC dataset has scene aggregation characteristics. By exploiting these properties, we propose Scene Clustering Based Pseudo-labeling Strategy (SCP-Label), a simple yet effective method to employ in post-processing. The SCP-Label brings greater accuracy by assigning the same label to objects within the same scene while also mitigating bias and confusion with model ensembles. Its performance surpasses the official baseline by a large margin of +20.57% Accuracy on Track 1 (SAR), and +31.86% Accuracy on Track 2 (SAR+EO), demonstrating the potential of SCP-Label as post-processing. Finally, we win the championship both on Track1 and Track2 in the CVPR 2022 Perception Beyond the Visible Spectrum (PBVS) Workshop MAVOC Challenge. Our code is available at https://github.com/HowieChangchn/SCP-Label.

Zhengyao Fang, Zexi Jia, Yijia Zhong et al.

Recent advances in text-to-image (T2I) generation have greatly improved visual quality, yet producing images that appear visually authentic to real-world photography remains challenging. This is partly due to biases in existing evaluation paradigms: human ratings and preference-trained metrics often favor visually vivid images with exaggerated saturation and contrast, which make generations often too vivid to be real even when prompted for realistic-style images. To address this issue, we present Color Fidelity Dataset (CFD) and Color Fidelity Metric (CFM) for objective evaluation of color fidelity in realistic-style generations. CFD contains over 1.3M real and synthetic images with ordered levels of color realism, while CFM employs a multimodal encoder to learn perceptual color fidelity. In addition, we propose a training-free Color Fidelity Refinement (CFR) that adaptively modulates spatial-temporal guidance scale in generation, thereby enhancing color authenticity. Together, CFD supports CFM for assessment, whose learned attention further guides CFR to refine T2I fidelity, forming a progressive framework for assessing and improving color fidelity in realistic-style T2I generation. The dataset and code are available at https://github.com/ZhengyaoFang/CFM.

Jaime Spencer, C. Stella Qian, Michaela Trescakova et al.

This paper discusses the results for the second edition of the Monocular Depth Estimation Challenge (MDEC). This edition was open to methods using any form of supervision, including fully-supervised, self-supervised, multi-task or proxy depth. The challenge was based around the SYNS-Patches dataset, which features a wide diversity of environments with high-quality dense ground-truth. This includes complex natural environments, e.g. forests or fields, which are greatly underrepresented in current benchmarks. The challenge received eight unique submissions that outperformed the provided SotA baseline on any of the pointcloud- or image-based metrics. The top supervised submission improved relative F-Score by 27.62%, while the top self-supervised improved it by 16.61%. Supervised submissions generally leveraged large collections of datasets to improve data diversity. Self-supervised submissions instead updated the network architecture and pretrained backbones. These results represent a significant progress in the field, while highlighting avenues for future research, such as reducing interpolation artifacts at depth boundaries, improving self-supervised indoor performance and overall natural image accuracy.

Lingdong Kong, Yaru Niu, Shaoyuan Xie et al.

Accurate depth estimation under out-of-distribution (OoD) scenarios, such as adverse weather conditions, sensor failure, and noise contamination, is desirable for safety-critical applications. Existing depth estimation systems, however, suffer inevitably from real-world corruptions and perturbations and are struggled to provide reliable depth predictions under such cases. In this paper, we summarize the winning solutions from the RoboDepth Challenge -- an academic competition designed to facilitate and advance robust OoD depth estimation. This challenge was developed based on the newly established KITTI-C and NYUDepth2-C benchmarks. We hosted two stand-alone tracks, with an emphasis on robust self-supervised and robust fully-supervised depth estimation, respectively. Out of more than two hundred participants, nine unique and top-performing solutions have appeared, with novel designs ranging from the following aspects: spatial- and frequency-domain augmentations, masked image modeling, image restoration and super-resolution, adversarial training, diffusion-based noise suppression, vision-language pre-training, learned model ensembling, and hierarchical feature enhancement. Extensive experimental analyses along with insightful observations are drawn to better understand the rationale behind each design. We hope this challenge could lay a solid foundation for future research on robust and reliable depth estimation and beyond. The datasets, competition toolkit, workshop recordings, and source code from the winning teams are publicly available on the challenge website.

Hui Kang, Sheng Liu, Huaxi Huang et al.

In recent years, research on learning with noisy labels has focused on devising novel algorithms that can achieve robustness to noisy training labels while generalizing to clean data. These algorithms often incorporate sophisticated techniques, such as noise modeling, label correction, and co-training. In this study, we demonstrate that a simple baseline using cross-entropy loss, combined with widely used regularization strategies like learning rate decay, model weights average, and data augmentations, can outperform state-of-the-art methods. Our findings suggest that employing a combination of regularization strategies can be more effective than intricate algorithms in tackling the challenges of learning with noisy labels. While some of these regularization strategies have been utilized in previous noisy label learning research, their full potential has not been thoroughly explored. Our results encourage a reevaluation of benchmarks for learning with noisy labels and prompt reconsideration of the role of specialized learning algorithms designed for training with noisy labels.

Chaojian Yu, Dawei Zhou, Li Shen et al.

Adversarial training (AT) is proved to reliably improve network's robustness against adversarial data. However, current AT with a pre-specified perturbation budget has limitations in learning a robust network. Firstly, applying a pre-specified perturbation budget on networks of various model capacities will yield divergent degree of robustness disparity between natural and robust accuracies, which deviates from robust network's desideratum. Secondly, the attack strength of adversarial training data constrained by the pre-specified perturbation budget fails to upgrade as the growth of network robustness, which leads to robust overfitting and further degrades the adversarial robustness. To overcome these limitations, we propose \emph{Strength-Adaptive Adversarial Training} (SAAT). Specifically, the adversary employs an adversarial loss constraint to generate adversarial training data. Under this constraint, the perturbation budget will be adaptively adjusted according to the training state of adversarial data, which can effectively avoid robust overfitting. Besides, SAAT explicitly constrains the attack strength of training data through the adversarial loss, which manipulates model capacity scheduling during training, and thereby can flexibly control the degree of robustness disparity and adjust the tradeoff between natural accuracy and robustness. Extensive experiments show that our proposal boosts the robustness of adversarial training.

Yuhao Wu, Xiaobo Xia, Jun Yu et al.

Training a classifier exploiting a huge amount of supervised data is expensive or even prohibited in a situation, where the labeling cost is high. The remarkable progress in working with weaker forms of supervision is binary classification from multiple unlabeled datasets which requires the knowledge of exact class priors for all unlabeled datasets. However, the availability of class priors is restrictive in many real-world scenarios. To address this issue, we propose to solve a new problem setting, i.e., binary classification from multiple unlabeled datasets with only one pairwise numerical relationship of class priors (MU-OPPO), which knows the relative order (which unlabeled dataset has a higher proportion of positive examples) of two class-prior probabilities for two datasets among multiple unlabeled datasets. In MU-OPPO, we do not need the class priors for all unlabeled datasets, but we only require that there exists a pair of unlabeled datasets for which we know which unlabeled dataset has a larger class prior. Clearly, this form of supervision is easier to be obtained, which can make labeling costs almost free. We propose a novel framework to handle the MU-OPPO problem, which consists of four sequential modules: (i) pseudo label assignment; (ii) confident example collection; (iii) class prior estimation; (iv) classifier training with estimated class priors. Theoretically, we analyze the gap between estimated class priors and true class priors under the proposed framework. Empirically, we confirm the superiority of our framework with comprehensive experiments. Experimental results demonstrate that our framework brings smaller estimation errors of class priors and better performance of binary classification.

Xuan Li, Yining Wang, Haocai Luo et al.

Retrieval-Augmented Generation (RAG) has become a pivotal paradigm for Large Language Models (LLMs), yet current approaches struggle with visually rich documents by treating text and images as isolated retrieval targets. Existing methods relying solely on cosine similarity often fail to capture the semantic reinforcement provided by cross-modal alignment and layout-induced coherence. To address these limitations, we propose BayesRAG, a novel multimodal retrieval framework grounded in Bayesian inference and Dempster-Shafer evidence theory. Unlike traditional approaches that rank candidates strictly by similarity, BayesRAG models the intrinsic consistency of retrieved candidates across modalities as probabilistic evidence to refine retrieval confidence. Specifically, our method computes the posterior association probability for combinations of multimodal retrieval results, prioritizing text-image pairs that mutually corroborate each other in terms of both semantics and layout. Extensive experiments demonstrate that BayesRAG significantly outperforms state-of-the-art (SOTA) methods on challenging multimodal benchmarks. This study establishes a new paradigm for multimodal retrieval fusion that effectively resolves the isolation of heterogeneous modalities through an evidence fusion mechanism and enhances the robustness of retrieval outcomes. Our code is available at https://github.com/TioeAre/BayesRAG.

Jianjian Qin, Chunzhi Gu, Jun Yu et al.

We present PD-REAL, a novel large-scale dataset for unsupervised anomaly detection (AD) in the 3D domain. It is motivated by the fact that 2D-only representations in the AD task may fail to capture the geometric structures of anomalies due to uncertainty in lighting conditions or shooting angles. PD-REAL consists entirely of Play-Doh models for 15 object categories and focuses on the analysis of potential benefits from 3D information in a controlled environment. Specifically, objects are first created with six types of anomalies, such as dent, crack, or perforation, and then photographed under different lighting conditions to mimic real-world inspection scenarios. To demonstrate the usefulness of 3D information, we use a commercially available RealSense camera to capture RGB and depth images. Compared to the existing 3D dataset for AD tasks, the data acquisition of PD-REAL is significantly cheaper, easily scalable and easier to control variables. Extensive evaluations with state-of-the-art AD algorithms on our dataset demonstrate the benefits as well as challenges of using 3D information. Our dataset can be downloaded from https://github.com/Andy-cs008/PD-REAL

Zhou Yu, Zitian Jin, Jun Yu et al.

Recent advances in Transformer architectures [1] have brought remarkable improvements to visual question answering (VQA). Nevertheless, Transformer-based VQA models are usually deep and wide to guarantee good performance, so they can only run on powerful GPU servers and cannot run on capacity-restricted platforms such as mobile phones. Therefore, it is desirable to learn an elastic VQA model that supports adaptive pruning at runtime to meet the efficiency constraints of different platforms. To this end, we present the bilaterally slimmable Transformer (BST), a general framework that can be seamlessly integrated into arbitrary Transformer-based VQA models to train a single model once and obtain various slimmed submodels of different widths and depths. To verify the effectiveness and generality of this method, we integrate the proposed BST framework with three typical Transformer-based VQA approaches, namely MCAN [2], UNITER [3], and CLIP-ViL [4], and conduct extensive experiments on two commonly-used benchmark datasets. In particular, one slimmed MCAN-BST submodel achieves comparable accuracy on VQA-v2, while being 0.38x smaller in model size and having 0.27x fewer FLOPs than the reference MCAN model. The smallest MCAN-BST submodel only has 9M parameters and 0.16G FLOPs during inference, making it possible to deploy it on a mobile device with less than 60 ms latency.

Buyu Liu, Kai Wang, Yansong Liu et al.

This work aims to address the multi-view perspective RGB generation from text prompts given Bird-Eye-View(BEV) semantics. Unlike prior methods that neglect layout consistency, lack the ability to handle detailed text prompts, or are incapable of generalizing to unseen view points, MVPbev simultaneously generates cross-view consistent images of different perspective views with a two-stage design, allowing object-level control and novel view generation at test-time. Specifically, MVPbev firstly projects given BEV semantics to perspective view with camera parameters, empowering the model to generalize to unseen view points. Then we introduce a multi-view attention module where special initialization and de-noising processes are introduced to explicitly enforce local consistency among overlapping views w.r.t. cross-view homography. Last but not least, MVPbev further allows test-time instance-level controllability by refining a pre-trained text-to-image diffusion model. Our extensive experiments on NuScenes demonstrate that our method is capable of generating high-resolution photorealistic images from text descriptions with thousands of training samples, surpassing the state-of-the-art methods under various evaluation metrics. We further demonstrate the advances of our method in terms of generalizability and controllability with the help of novel evaluation metrics and comprehensive human analysis. Our code, data, and model can be found in \url{https://github.com/kkaiwwana/MVPbev}.

Fang Gao, XueTao Li, Jun Yu et al.

The advent of Chat-GPT has led to a surge of interest in Embodied AI. However, many existing Embodied AI models heavily rely on massive interactions with training environments, which may not be practical in real-world situations. To this end, the Maniskill2 has introduced a full-physics simulation benchmark for manipulating various 3D objects. This benchmark enables agents to be trained using diverse datasets of demonstrations and evaluates their ability to generalize to unseen scenarios in testing environments. In this paper, we propose a novel two-stage fine-tuning strategy that aims to further enhance the generalization capability of our model based on the Maniskill2 benchmark. Through extensive experiments, we demonstrate the effectiveness of our approach by achieving the 1st prize in all three tracks of the ManiSkill2 Challenge. Our findings highlight the potential of our method to improve the generalization abilities of Embodied AI models and pave the way for their ractical applications in real-world scenarios. All codes and models of our solution is available at https://github.com/xtli12/GXU-LIPE.git

Zhuo Huang, Li Shen, Jun Yu et al.

Semi-Supervised Learning (SSL) has been an effective way to leverage abundant unlabeled data with extremely scarce labeled data. However, most SSL methods are commonly based on instance-wise consistency between different data transformations. Therefore, the label guidance on labeled data is hard to be propagated to unlabeled data. Consequently, the learning process on labeled data is much faster than on unlabeled data which is likely to fall into a local minima that does not favor unlabeled data, leading to sub-optimal generalization performance. In this paper, we propose FlatMatch which minimizes a cross-sharpness measure to ensure consistent learning performance between the two datasets. Specifically, we increase the empirical risk on labeled data to obtain a worst-case model which is a failure case that needs to be enhanced. Then, by leveraging the richness of unlabeled data, we penalize the prediction difference (i.e., cross-sharpness) between the worst-case model and the original model so that the learning direction is beneficial to generalization on unlabeled data. Therefore, we can calibrate the learning process without being limited to insufficient label information. As a result, the mismatched learning performance can be mitigated, further enabling the effective exploitation of unlabeled data and improving SSL performance. Through comprehensive validation, we show FlatMatch achieves state-of-the-art results in many SSL settings.

Muyang Li, Runze Wu, Haoyu Liu et al.

Semi-supervised learning (SSL) has been a fundamental challenge in machine learning for decades. The primary family of SSL algorithms, known as pseudo-labeling, involves assigning pseudo-labels to confident unlabeled instances and incorporating them into the training set. Therefore, the selection criteria of confident instances are crucial to the success of SSL. Recently, there has been growing interest in the development of SSL methods that use dynamic or adaptive thresholds. Yet, these methods typically apply the same threshold to all samples, or use class-dependent thresholds for instances belonging to a certain class, while neglecting instance-level information. In this paper, we propose the study of instance-dependent thresholds, which has the highest degree of freedom compared with existing methods. Specifically, we devise a novel instance-dependent threshold function for all unlabeled instances by utilizing their instance-level ambiguity and the instance-dependent error rates of pseudo-labels, so instances that are more likely to have incorrect pseudo-labels will have higher thresholds. Furthermore, we demonstrate that our instance-dependent threshold function provides a bounded probabilistic guarantee for the correctness of the pseudo-labels it assigns.

Zhuo Huang, Muyang Li, Li Shen et al.

Out-of-Distribution (OOD) Generalization aims to learn robust models that generalize well to various environments without fitting to distribution-specific features. Recent studies based on Lottery Ticket Hypothesis (LTH) address this problem by minimizing the learning target to find some of the parameters that are critical to the task. However, in OOD problems, such solutions are suboptimal as the learning task contains severe distribution noises, which can mislead the optimization process. Therefore, apart from finding the task-related parameters (i.e., invariant parameters), we propose Exploring Variant parameters for Invariant Learning (EVIL) which also leverages the distribution knowledge to find the parameters that are sensitive to distribution shift (i.e., variant parameters). Once the variant parameters are left out of invariant learning, a robust subnetwork that is resistant to distribution shift can be found. Additionally, the parameters that are relatively stable across distributions can be considered invariant ones to improve invariant learning. By fully exploring both variant and invariant parameters, our EVIL can effectively identify a robust subnetwork to improve OOD generalization. In extensive experiments on integrated testbed: DomainBed, EVIL can effectively and efficiently enhance many popular methods, such as ERM, IRM, SAM, etc.

Haifeng Wang, Hua Wu, Tian Wu et al.

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

Ye Yu, Fengxin Chen, Jun Yu et al.

While recent low-light image enhancement (LLIE) methods have made significant advancements, they still face challenges in terms of low visual quality and weak generalization ability when applied to complex scenarios. To address these issues, we propose a semi-supervised method based on latent mean-teacher and Gaussian process, named LMT-GP. We first design a latent mean-teacher framework that integrates both labeled and unlabeled data, as well as their latent vectors, into model training. Meanwhile, we use a mean-teacher-assisted Gaussian process learning strategy to establish a connection between the latent and pseudo-latent vectors obtained from the labeled and unlabeled data. To guide the learning process, we utilize an assisted Gaussian process regression (GPR) loss function. Furthermore, we design a pseudo-label adaptation module (PAM) to ensure the reliability of the network learning. To demonstrate our method's generalization ability and effectiveness, we apply it to multiple LLIE datasets and high-level vision tasks. Experiment results demonstrate that our method achieves high generalization performance and image quality. The code is available at https://github.com/HFUT-CV/LMT-GP.

Zijian Zhang, Zhou Zhao, Jun Yu et al.

Diffusion models have recently exhibited remarkable abilities to synthesize striking image samples since the introduction of denoising diffusion probabilistic models (DDPMs). Their key idea is to disrupt images into noise through a fixed forward process and learn its reverse process to generate samples from noise in a denoising way. For conditional DDPMs, most existing practices relate conditions only to the reverse process and fit it to the reversal of unconditional forward process. We find this will limit the condition modeling and generation in a small time window. In this paper, we propose a novel and flexible conditional diffusion model by introducing conditions into the forward process. We utilize extra latent space to allocate an exclusive diffusion trajectory for each condition based on some shifting rules, which will disperse condition modeling to all timesteps and improve the learning capacity of model. We formulate our method, which we call \textbf{ShiftDDPMs}, and provide a unified point of view on existing related methods. Extensive qualitative and quantitative experiments on image synthesis demonstrate the feasibility and effectiveness of ShiftDDPMs.

Jianjian Qin, Chunzhi Gu, Jun Yu et al.

Anomaly detection, which is a critical and popular topic in computer vision, aims to detect anomalous samples that are different from the normal (i.e., non-anomalous) ones. The current mainstream methods focus on anomaly detection for images, whereas little attention has been paid to 3D point cloud. In this paper, drawing inspiration from the knowledge transfer ability of teacher-student architecture and the impressive feature extraction capability of recent neural networks, we design a teacher-student structured model for 3D anomaly detection. Specifically, we use feature space alignment, dimension zoom, and max pooling to extract the features of the point cloud and then minimize a multi-scale loss between the feature vectors produced by the teacher and the student networks. Moreover, our method only requires very few normal samples to train the student network due to the teacher-student distillation mechanism. Once trained, the teacher-student network pair can be leveraged jointly to fulfill 3D point cloud anomaly detection based on the calculated anomaly score. For evaluation, we compare our method against the reconstruction-based method on the ShapeNet-Part dataset. The experimental results and ablation studies quantitatively and qualitatively confirm that our model can achieve higher performance compared with the state of the arts in 3D anomaly detection with very few training samples.

Mengyu Li, Jun Yu, Hongteng Xu et al.

As a valid metric of metric-measure spaces, Gromov-Wasserstein (GW) distance has shown the potential for matching problems of structured data like point clouds and graphs. However, its application in practice is limited due to the high computational complexity. To overcome this challenge, we propose a novel importance sparsification method, called \textsc{Spar-GW}, to approximate GW distance efficiently. In particular, instead of considering a dense coupling matrix, our method leverages a simple but effective sampling strategy to construct a sparse coupling matrix and update it with few computations. The proposed \textsc{Spar-GW} method is applicable to the GW distance with arbitrary ground cost, and it reduces the complexity from $O(n^4)$ to $O(n^{2+δ})$ for an arbitrary small $δ>0$. Theoretically, the convergence and consistency of the proposed estimation for GW distance are established under mild regularity conditions. In addition, this method can be extended to approximate the variants of GW distance, including the entropic GW distance, the fused GW distance, and the unbalanced GW distance. Experiments show the superiority of our \textsc{Spar-GW} to state-of-the-art methods in both synthetic and real-world tasks.

Tao Li, Cheng Meng, Hongteng Xu et al.

Distribution comparison plays a central role in many machine learning tasks like data classification and generative modeling. In this study, we propose a novel metric, called Hilbert curve projection (HCP) distance, to measure the distance between two probability distributions with low complexity. In particular, we first project two high-dimensional probability distributions using Hilbert curve to obtain a coupling between them, and then calculate the transport distance between these two distributions in the original space, according to the coupling. We show that HCP distance is a proper metric and is well-defined for probability measures with bounded supports. Furthermore, we demonstrate that the modified empirical HCP distance with the $L_p$ cost in the $d$-dimensional space converges to its population counterpart at a rate of no more than $O(n^{-1/2\max\{d,p\}})$. To suppress the curse-of-dimensionality, we also develop two variants of the HCP distance using (learnable) subspace projections. Experiments on both synthetic and real-world data show that our HCP distance works as an effective surrogate of the Wasserstein distance with low complexity and overcomes the drawbacks of the sliced Wasserstein distance.

Mengyu Li, Jun Yu, Tao Li et al.

Sinkhorn algorithm has been used pervasively to approximate the solution to optimal transport (OT) and unbalanced optimal transport (UOT) problems. However, its practical application is limited due to the high computational complexity. To alleviate the computational burden, we propose a novel importance sparsification method, called Spar-Sink, to efficiently approximate entropy-regularized OT and UOT solutions. Specifically, our method employs natural upper bounds for unknown optimal transport plans to establish effective sampling probabilities, and constructs a sparse kernel matrix to accelerate Sinkhorn iterations, reducing the computational cost of each iteration from $O(n^2)$ to $\widetilde{O}(n)$ for a sample of size $n$. Theoretically, we show the proposed estimators for the regularized OT and UOT problems are consistent under mild regularity conditions. Experiments on various synthetic data demonstrate Spar-Sink outperforms mainstream competitors in terms of both estimation error and speed. A real-world echocardiogram data analysis shows Spar-Sink can effectively estimate and visualize cardiac cycles, from which one can identify heart failure and arrhythmia. To evaluate the numerical accuracy of cardiac cycle prediction, we consider the task of predicting the end-systole time point using the end-diastole one. Results show Spar-Sink performs as well as the classical Sinkhorn algorithm, requiring significantly less computational time.

Haodong Chen, Qiang Huang, Jiaqi Zhao et al.

Vision-Language Models (VLMs) are increasingly deployed in socially consequential settings, raising concerns about social bias driven by demographic cues. A central challenge in measuring such social bias is attribution under visual confounding: real-world images entangle race and gender with correlated factors such as background and clothing, obscuring attribution. We propose a \textbf{face-only counterfactual evaluation paradigm} that isolates demographic effects while preserving real-image realism. Starting from real photographs, we generate counterfactual variants by editing only facial attributes related to race and gender, keeping all other visual factors fixed. Based on this paradigm, we construct \textbf{FOCUS}, a dataset of 480 scene-matched counterfactual images across six occupations and ten demographic groups, and propose \textbf{REFLECT}, a benchmark comprising three decision-oriented tasks: two-alternative forced choice, multiple-choice socioeconomic inference, and numeric salary recommendation. Experiments on five state-of-the-art VLMs reveal that demographic disparities persist under strict visual control and vary substantially across task formulations. These findings underscore the necessity of controlled, counterfactual audits and highlight task design as a critical factor in evaluating social bias in multimodal models.

Jun Yu, Zhongpeng Cai, Renda Li et al.

Facial Expression Recognition (FER) is an important task in computer vision and has wide applications in human-computer interaction, intelligent security, emotion analysis, and other fields. However, the limited size of FER datasets limits the generalization ability of expression recognition models, resulting in ineffective model performance. To address this problem, we propose a semi-supervised learning framework that utilizes unlabeled face data to train expression recognition models effectively. Our method uses a dynamic threshold module (\textbf{DTM}) that can adaptively adjust the confidence threshold to fully utilize the face recognition (FR) data to generate pseudo-labels, thus improving the model's ability to model facial expressions. In the ABAW5 EXPR task, our method achieved excellent results on the official validation set.

Jun Yu, Zhongpeng Cai, Peng He et al.

Analysis of human affect plays a vital role in human-computer interaction (HCI) systems. Due to the difficulty in capturing large amounts of real-life data, most of the current methods have mainly focused on controlled environments, which limit their application scenarios. To tackle this problem, we propose our solution based on the ensemble learning method. Specifically, we formulate the problem as a classification task, and then train several expression classification models with different types of backbones--ResNet, EfficientNet and InceptionNet. After that, the outputs of several models are fused via model ensemble method to predict the final results. Moreover, we introduce the multi-fold ensemble method to train and ensemble several models with the same architecture but different data distributions to enhance the performance of our solution. We conduct many experiments on the AffWild2 dataset of the ABAW2022 Challenge, and the results demonstrate the effectiveness of our solution.

Xuan Li, Yining Wang, Yuchen Liu et al.

Chain-of-thought (CoT) reasoning improves large language models (LLMs) on difficult tasks, but it also makes inference expensive because every intermediate step must be generated as a discrete token. Latent reasoning reduces visible token generation by propagating continuous states, yet replacing explicit derivations with latent computation can hurt tasks that require symbolic checking. We propose Latent-Then-Explicit Reasoning (LaTER), a two-stage paradigm that first performs bounded exploration in a continuous latent space and then switches to explicit CoT for verification and answer generation. In a training-free instantiation, LaTER projects final-layer hidden states back to the input embedding space, preserves the latent KV cache, and uses entropy and model-native stop-token probes to decide when to switch. We find that strong reasoning models already exhibit structured latent trajectories under this interface. On Qwen3-14B, training-free LaTER reduces total token usage by 16%-32% on several benchmarks while matching or improving accuracy on most of them; for example, it improves AIME 2025 from 70.0% to 73.3% while reducing tokens from 15,730 to 10,661. We further construct Latent-Switch-69K, a supervised corpus that pairs condensed solution intuitions with shortened explicit derivations. Fine-tuning with latent rollout and halting supervision yields additional gains: trained LaTER reaches 80.0% accuracy on AIME 2025, 10.0 points above the standard CoT baseline, while using 33% fewer tokens. Our code, data, and model are available at https://github.com/TioeAre/LaTER.

Jun Yu, Renda Li, Zhongpeng Cai et al.

Human affective behavior analysis plays a vital role in human-computer interaction (HCI) systems. In this paper, we introduce our submission to the CVPR 2023 Competition on Affective Behavior Analysis in-the-wild (ABAW). We propose a single-stage trained AU detection framework. Specifically, in order to effectively extract facial local region features related to AU detection, we use a local region perception module to effectively extract features of different AUs. Meanwhile, we use a graph neural network-based relational learning module to capture the relationship between AUs. In addition, considering the role of the overall feature of the target face on AU detection, we also use the feature fusion module to fuse the feature information extracted by the backbone network and the AU feature information extracted by the relationship learning module. We also adopted some sampling methods, data augmentation techniques and post-processing strategies to further improve the performance of the model.

Chunzhi Gu, Jun Yu, Chao Zhang

Generative model-based motion prediction techniques have recently realized predicting controlled human motions, such as predicting multiple upper human body motions with similar lower-body motions. However, to achieve this, the state-of-the-art methods require either subsequently learning mapping functions to seek similar motions or training the model repetitively to enable control over the desired portion of body. In this paper, we propose a novel framework to learn disentangled representations for controllable human motion prediction. Our network involves a conditional variational auto-encoder (CVAE) architecture to model full-body human motion, and an extra CVAE path to learn only the corresponding partial-body (e.g., lower-body) motion. Specifically, the inductive bias imposed by the extra CVAE path encourages two latent variables in two paths to respectively govern separate representations for each partial-body motion. With a single training, our model is able to provide two types of controls for the generated human motions: (i) strictly controlling one portion of human body and (ii) adaptively controlling the other portion, by sampling from a pair of latent spaces. Additionally, we extend and adapt a sampling strategy to our trained model to diversify the controllable predictions. Our framework also potentially allows new forms of control by flexibly customizing the input for the extra CVAE path. Extensive experimental results and ablation studies demonstrate that our approach is capable of predicting state-of-the-art controllable human motions both qualitatively and quantitatively.

Chaojian Yu, Xiaolong Shi, Jun Yu et al.

Adversarial training (AT) constructs robust neural networks by incorporating adversarial perturbations into natural data. However, it is plagued by the issue of robust overfitting (RO), which severely damages the model's robustness. In this paper, we investigate RO from a novel feature generalization perspective. Specifically, we design factor ablation experiments to assess the respective impacts of natural data and adversarial perturbations on RO, identifying that the inducing factor of RO stems from natural data. Given that the only difference between adversarial and natural training lies in the inclusion of adversarial perturbations, we further hypothesize that adversarial perturbations degrade the generalization of features in natural data and verify this hypothesis through extensive experiments. Based on these findings, we provide a holistic view of RO from the feature generalization perspective and explain various empirical behaviors associated with RO. To examine our feature generalization perspective, we devise two representative methods, attack strength and data augmentation, to prevent the feature generalization degradation during AT. Extensive experiments conducted on benchmark datasets demonstrate that the proposed methods can effectively mitigate RO and enhance adversarial robustness.

Jun Yu, Jichao Zhu, Wangyuan Zhu et al.

Emotional Reaction Intensity(ERI) estimation is an important task in multimodal scenarios, and has fundamental applications in medicine, safe driving and other fields. In this paper, we propose a solution to the ERI challenge of the fifth Affective Behavior Analysis in-the-wild(ABAW), a dual-branch based multi-output regression model. The spatial attention is used to better extract visual features, and the Mel-Frequency Cepstral Coefficients technology extracts acoustic features, and a method named modality dropout is added to fusion multimodal features. Our method achieves excellent results on the official validation set.

Jun Yu, Shenshen Du, Guochen Xie et al.

Synthetic Aperture Radar (SAR) to electro-optical (EO) image translation is a fundamental task in remote sensing that can enrich the dataset by fusing information from different sources. Recently, many methods have been proposed to tackle this task, but they are still difficult to complete the conversion from low-resolution images to high-resolution images. Thus, we propose a framework, SAR2EO, aiming at addressing this challenge. Firstly, to generate high-quality EO images, we adopt the coarse-to-fine generator, multi-scale discriminators, and improved adversarial loss in the pix2pixHD model to increase the synthesis quality. Secondly, we introduce a denoising module to remove the noise in SAR images, which helps to suppress the noise while preserving the structural information of the images. To validate the effectiveness of the proposed framework, we conduct experiments on the dataset of the Multi-modal Aerial View Imagery Challenge (MAVIC), which consists of large-scale SAR and EO image pairs. The experimental results demonstrate the superiority of our proposed framework, and we win the first place in the MAVIC held in CVPR PBVS 2023.

Jun Yu, Yutong Dai, Xiaokang Liu et al.

MTL is a learning paradigm that effectively leverages both task-specific and shared information to address multiple related tasks simultaneously. In contrast to STL, MTL offers a suite of benefits that enhance both the training process and the inference efficiency. MTL's key advantages encompass streamlined model architecture, performance enhancement, and cross-domain generalizability. Over the past twenty years, MTL has become widely recognized as a flexible and effective approach in various fields, including CV, NLP, recommendation systems, disease prognosis and diagnosis, and robotics. This survey provides a comprehensive overview of the evolution of MTL, encompassing the technical aspects of cutting-edge methods from traditional approaches to deep learning and the latest trend of pretrained foundation models. Our survey methodically categorizes MTL techniques into five key areas: regularization, relationship learning, feature propagation, optimization, and pre-training. This categorization not only chronologically outlines the development of MTL but also dives into various specialized strategies within each category. Furthermore, the survey reveals how the MTL evolves from handling a fixed set of tasks to embracing a more flexible approach free from task or modality constraints. It explores the concepts of task-promptable and -agnostic training, along with the capacity for ZSL, which unleashes the untapped potential of this historically coveted learning paradigm. Overall, we hope this survey provides the research community with a comprehensive overview of the advancements in MTL from its inception in 1997 to the present in 2023. We address present challenges and look ahead to future possibilities, shedding light on the opportunities and potential avenues for MTL research in a broad manner. This project is publicly available at https://github.com/junfish/Awesome-Multitask-Learning.

Zhenhua Ning, Xin Li, Jun Yu et al.

While 3D Gaussian Splatting (3DGS) has demonstrated impressive real-time rendering performance, its efficacy remains constrained by a reliance on heuristic density control. Despite numerous refinements to these handcrafted rules, such methods inherently lack the flexibility to adapt to diverse scenes with complex geometries. In this paper, we propose a paradigm shift for density control from rigid heuristics to fully learnable policies. Specifically, we introduce \textbf{LeGS}, a framework that reformulates density control as a parameterized policy network optimized via Reinforcement Learning (RL). Central to our approach is the tailored effective reward function grounded in sensitivity analysis, which precisely quantifies the marginal contribution of individual Gaussians to reconstruction quality. To maintain computational tractability, we derive a closed-form solution that reduces the complexity of reward calculation from $O(N^2)$ to $O(N)$. Extensive experiments on the Mip-NeRF 360, Tanks \& Temples, and Deep Blending datasets demonstrate that \textbf{LeGS} significantly outperforms state-of-the-art methods, striking a superior balance between reconstruction quality and efficiency. The code will be released at https://github.com/AaronNZH/LeGS

Junjie Zhang, Feng Zhao, Hanqiang Liu et al.

The booming remote sensing (RS) technology is giving rise to a novel multimodality generalization task, which requires the model to overcome data heterogeneity while possessing powerful cross-scene generalization ability. Moreover, most vision-language models (VLMs) usually describe surface materials in RS images using universal texts, lacking proprietary linguistic prior knowledge specific to different RS vision modalities. In this work, we formalize RS multimodality generalization (RSMG) as a learning paradigm, and propose a frequency-aware vision-language multimodality generalization network (FVMGN) for RS image classification. Specifically, a diffusion-based training-test-time augmentation (DTAug) strategy is designed to reconstruct multimodal land-cover distributions, enriching input information for FVMGN. Following that, to overcome multimodal heterogeneity, a multimodal wavelet disentanglement (MWDis) module is developed to learn cross-domain invariant features by resampling low and high frequency components in the frequency domain. Considering the characteristics of RS vision modalities, shared and proprietary class texts is designed as linguistic inputs for the transformer-based text encoder to extract diverse text features. For multimodal vision inputs, a spatial-frequency-aware image encoder (SFIE) is constructed to realize local-global feature reconstruction and representation. Finally, a multiscale spatial-frequency feature alignment (MSFFA) module is suggested to construct a unified semantic space, ensuring refined multiscale alignment of different text and vision features in spatial and frequency domains. Extensive experiments show that FVMGN has the excellent multimodality generalization ability compared with state-of-the-art (SOTA) methods.

Xiaoxing You, Qiang Huang, Lingyu Li et al.

Multimodal Summarization (MMS) aims to generate concise textual summaries by understanding and integrating information across videos, transcripts, and images. However, existing approaches still suffer from three main challenges: (1) reliance on domain-specific supervision, (2) implicit fusion with weak cross-modal grounding, and (3) flat temporal modeling without event transitions. To address these issues, we introduce **CoE**, a training-free MMS framework that performs structured reasoning through a **Chain-of-Events** guided by a Hierarchical Event Graph (HEG). The HEG encodes textual semantics into an explicit event hierarchy that scaffolds cross-modal grounding and temporal reasoning. Guided by this structure, **CoE** localizes key visual cues, models event evolution and causal transitions, and refines outputs via lightweight style adaptation for domain alignment. Extensive experiments on eight diverse datasets demonstrate that **CoE** consistently outperforms state-of-the-art video CoT baselines, achieving average gains of **+3.04 ROUGE**, **+9.51 CIDEr**, and **+1.88 BERTScore**, highlighting its robustness, interpretability, and cross-domain generalization. Our code is available at https://github.com/youxiaoxing/CoE.

Xiangyang Miao, Guobao Xiao, Shiping Wang et al.

Correspondence pruning aims to establish reliable correspondences between two related images and recover relative camera motion. Existing approaches often employ a progressive strategy to handle the local and global contexts, with a prominent emphasis on transitioning from local to global, resulting in the neglect of interactions between different contexts. To tackle this issue, we propose a parallel context learning strategy that involves acquiring bilateral consensus for the two-view correspondence pruning task. In our approach, we design a distinctive self-attention block to capture global context and parallel process it with the established local context learning module, which enables us to simultaneously capture both local and global consensuses. By combining these local and global consensuses, we derive the required bilateral consensus. We also design a recalibration block, reducing the influence of erroneous consensus information and enhancing the robustness of the model. The culmination of our efforts is the Bilateral Consensus Learning Network (BCLNet), which efficiently estimates camera pose and identifies inliers (true correspondences). Extensive experiments results demonstrate that our network not only surpasses state-of-the-art methods on benchmark datasets but also showcases robust generalization abilities across various feature extraction techniques. Noteworthily, BCLNet obtains 3.98\% mAP5$^{\circ}$ gains over the second best method on unknown outdoor dataset, and obviously accelerates model training speed. The source code will be available at: https://github.com/guobaoxiao/BCLNet.

Jianjian Qin, Chunzhi Gu, Jun Yu et al.

Anomaly detection (AD) is a fundamental task in computer vision. It aims to identify incorrect image data patterns which deviate from the normal ones. Conventional methods generally address AD by preparing augmented negative samples to enforce self-supervised learning. However, these techniques typically do not consider semantics during augmentation, leading to the generation of unrealistic or invalid negative samples. Consequently, the feature extraction network can be hindered from embedding critical features. In this study, inspired by visual attention learning approaches, we propose CutSwap, which leverages saliency guidance to incorporate semantic cues for augmentation. Specifically, we first employ LayerCAM to extract multilevel image features as saliency maps and then perform clustering to obtain multiple centroids. To fully exploit saliency guidance, on each map, we select a pixel pair from the cluster with the highest centroid saliency to form a patch pair. Such a patch pair includes highly similar context information with dense semantic correlations. The resulting negative sample is created by swapping the locations of the patch pair. Compared to prior augmentation methods, CutSwap generates more subtle yet realistic negative samples to facilitate quality feature learning. Extensive experimental and ablative evaluations demonstrate that our method achieves state-of-the-art AD performance on two mainstream AD benchmark datasets.

Junwen Guo, Guobao Xiao, Shiping Wang et al.

Most of existing correspondence pruning methods only concentrate on gathering the context information as much as possible while neglecting effective ways to utilize such information. In order to tackle this dilemma, in this paper we propose Graph Context Transformation Network (GCT-Net) enhancing context information to conduct consensus guidance for progressive correspondence pruning. Specifically, we design the Graph Context Enhance Transformer which first generates the graph network and then transforms it into multi-branch graph contexts. Moreover, it employs self-attention and cross-attention to magnify characteristics of each graph context for emphasizing the unique as well as shared essential information. To further apply the recalibrated graph contexts to the global domain, we propose the Graph Context Guidance Transformer. This module adopts a confident-based sampling strategy to temporarily screen high-confidence vertices for guiding accurate classification by searching global consensus between screened vertices and remaining ones. The extensive experimental results on outlier removal and relative pose estimation clearly demonstrate the superior performance of GCT-Net compared to state-of-the-art methods across outdoor and indoor datasets. The source code will be available at: https://github.com/guobaoxiao/GCT-Net/.

Tingjun Dai, Mingfei Han, Tingwen Du et al.

Measurement of task progress through explicit, actionable milestones is critical for robust robotic manipulation. This progress awareness enables a model to ground its current task status, anticipate verifiable intermediate states, and detect and recover from failures when progress stalls. To embody this capability, we introduce See, Plan, Rewind (SPR), a progress-aware vision-language-action framework that dynamically grounds language instructions into a sequence of spatial subgoals. SPR operates through a continuous core cycle, Seeing the current state and upcoming milestone, Planning a trajectory towards the next 2D waypoint, and Rewinding to a recoverable state upon failure by monitoring progress against the expected sequence. This closed-loop approach enables robust error correction without requiring additional training data or auxiliary models. Extensive experiments demonstrate the framework's effectiveness, generalization and robustness: SPR outperforms the MolmoAct baseline by 5\% on the LIBERO benchmark. On the challenging LIBERO-Plus benchmark with unseen instructions and initial states, SPR achieves state-of-the-art robustness with the smallest performance drop, surpassing OpenVLA-OFT and UniVLA, demonstrating superior out-of-distribution robustness.

Guobao Xiao, Jun Yu, Jiayi Ma et al.

Estimating reliable geometric model parameters from the data with severe outliers is a fundamental and important task in computer vision. This paper attempts to sample high-quality subsets and select model instances to estimate parameters in the multi-structural data. To address this, we propose an effective method called Latent Semantic Consensus (LSC). The principle of LSC is to preserve the latent semantic consensus in both data points and model hypotheses. Specifically, LSC formulates the model fitting problem into two latent semantic spaces based on data points and model hypotheses, respectively. Then, LSC explores the distributions of points in the two latent semantic spaces, to remove outliers, generate high-quality model hypotheses, and effectively estimate model instances. Finally, LSC is able to provide consistent and reliable solutions within only a few milliseconds for general multi-structural model fitting, due to its deterministic fitting nature and efficiency. Compared with several state-of-the-art model fitting methods, our LSC achieves significant superiority for the performance of both accuracy and speed on synthetic data and real images. The code will be available at https://github.com/guobaoxiao/LSC.