Xilin Chen

Botao Ye, Hong Chang, Bingpeng Ma et al.

The current popular two-stream, two-stage tracking framework extracts the template and the search region features separately and then performs relation modeling, thus the extracted features lack the awareness of the target and have limited target-background discriminability. To tackle the above issue, we propose a novel one-stream tracking (OSTrack) framework that unifies feature learning and relation modeling by bridging the template-search image pairs with bidirectional information flows. In this way, discriminative target-oriented features can be dynamically extracted by mutual guidance. Since no extra heavy relation modeling module is needed and the implementation is highly parallelized, the proposed tracker runs at a fast speed. To further improve the inference efficiency, an in-network candidate early elimination module is proposed based on the strong similarity prior calculated in the one-stream framework. As a unified framework, OSTrack achieves state-of-the-art performance on multiple benchmarks, in particular, it shows impressive results on the one-shot tracking benchmark GOT-10k, i.e., achieving 73.7% AO, improving the existing best result (SwinTrack) by 4.3\%. Besides, our method maintains a good performance-speed trade-off and shows faster convergence. The code and models are available at https://github.com/botaoye/OSTrack.

Xinqian Gu, Hong Chang, Bingpeng Ma et al.

The key to address clothes-changing person re-identification (re-id) is to extract clothes-irrelevant features, e.g., face, hairstyle, body shape, and gait. Most current works mainly focus on modeling body shape from multi-modality information (e.g., silhouettes and sketches), but do not make full use of the clothes-irrelevant information in the original RGB images. In this paper, we propose a Clothes-based Adversarial Loss (CAL) to mine clothes-irrelevant features from the original RGB images by penalizing the predictive power of re-id model w.r.t. clothes. Extensive experiments demonstrate that using RGB images only, CAL outperforms all state-of-the-art methods on widely-used clothes-changing person re-id benchmarks. Besides, compared with images, videos contain richer appearance and additional temporal information, which can be used to model proper spatiotemporal patterns to assist clothes-changing re-id. Since there is no publicly available clothes-changing video re-id dataset, we contribute a new dataset named CCVID and show that there exists much room for improvement in modeling spatiotemporal information. The code and new dataset are available at: https://github.com/guxinqian/Simple-CCReID.

Dongyang Liu, Meina Kan, Shiguang Shan et al.

Feature distillation makes the student mimic the intermediate features of the teacher. Nearly all existing feature-distillation methods use L2 distance or its slight variants as the distance metric between teacher and student features. However, while L2 distance is isotropic w.r.t. all dimensions, the neural network's operation on different dimensions is usually anisotropic, i.e., perturbations with the same 2-norm but in different dimensions of intermediate features lead to changes in the final output with largely different magnitude. Considering this, we argue that the similarity between teacher and student features should not be measured merely based on their appearance (i.e., L2 distance), but should, more importantly, be measured by their difference in function, namely how later layers of the network will read, decode, and process them. Therefore, we propose Function-Consistent Feature Distillation (FCFD), which explicitly optimizes the functional similarity between teacher and student features. The core idea of FCFD is to make teacher and student features not only numerically similar, but more importantly produce similar outputs when fed to the later part of the same network. With FCFD, the student mimics the teacher more faithfully and learns more from the teacher. Extensive experiments on image classification and object detection demonstrate the superiority of FCFD to existing methods. Furthermore, we can combine FCFD with many existing methods to obtain even higher accuracy. Our codes are available at https://github.com/LiuDongyang6/FCFD.

Wenrui Liu, Hong Chang, Bingpeng Ma et al.

Reconstruction-based anomaly detection models achieve their purpose by suppressing the generalization ability for anomaly. However, diverse normal patterns are consequently not well reconstructed as well. Although some efforts have been made to alleviate this problem by modeling sample diversity, they suffer from shortcut learning due to undesired transmission of abnormal information. In this paper, to better handle the tradeoff problem, we propose Diversity-Measurable Anomaly Detection (DMAD) framework to enhance reconstruction diversity while avoid the undesired generalization on anomalies. To this end, we design Pyramid Deformation Module (PDM), which models diverse normals and measures the severity of anomaly by estimating multi-scale deformation fields from reconstructed reference to original input. Integrated with an information compression module, PDM essentially decouples deformation from prototypical embedding and makes the final anomaly score more reliable. Experimental results on both surveillance videos and industrial images demonstrate the effectiveness of our method. In addition, DMAD works equally well in front of contaminated data and anomaly-like normal samples.

Zhongqi Wang, Jie Zhang, Shiguang Shan et al.

While text-to-image diffusion models demonstrate impressive generation capabilities, they also exhibit vulnerability to backdoor attacks, which involve the manipulation of model outputs through malicious triggers. In this paper, for the first time, we propose a comprehensive defense method named T2IShield to detect, localize, and mitigate such attacks. Specifically, we find the "Assimilation Phenomenon" on the cross-attention maps caused by the backdoor trigger. Based on this key insight, we propose two effective backdoor detection methods: Frobenius Norm Threshold Truncation and Covariance Discriminant Analysis. Besides, we introduce a binary-search approach to localize the trigger within a backdoor sample and assess the efficacy of existing concept editing methods in mitigating backdoor attacks. Empirical evaluations on two advanced backdoor attack scenarios show the effectiveness of our proposed defense method. For backdoor sample detection, T2IShield achieves a detection F1 score of 88.9$\%$ with low computational cost. Furthermore, T2IShield achieves a localization F1 score of 86.4$\%$ and invalidates 99$\%$ poisoned samples. Codes are released at https://github.com/Robin-WZQ/T2IShield.

Ruibing Hou, Hong Chang, Bingpeng Ma et al.

Learning generalizable representation and classifier for class-imbalanced data is challenging for data-driven deep models. Most studies attempt to re-balance the data distribution, which is prone to overfitting on tail classes and underfitting on head classes. In this work, we propose Dual Compensation Residual Networks to better fit both tail and head classes. Firstly, we propose dual Feature Compensation Module (FCM) and Logit Compensation Module (LCM) to alleviate the overfitting issue. The design of these two modules is based on the observation: an important factor causing overfitting is that there is severe feature drift between training and test data on tail classes. In details, the test features of a tail category tend to drift towards feature cloud of multiple similar head categories. So FCM estimates a multi-mode feature drift direction for each tail category and compensate for it. Furthermore, LCM translates the deterministic feature drift vector estimated by FCM along intra-class variations, so as to cover a larger effective compensation space, thereby better fitting the test features. Secondly, we propose a Residual Balanced Multi-Proxies Classifier (RBMC) to alleviate the under-fitting issue. Motivated by the observation that re-balancing strategy hinders the classifier from learning sufficient head knowledge and eventually causes underfitting, RBMC utilizes uniform learning with a residual path to facilitate classifier learning. Comprehensive experiments on Long-tailed and Class-Incremental benchmarks validate the efficacy of our method.

Songtao Luo, Shuang Yang, Shiguang Shan et al.

In this paper, we propose a novel method for speaker adaptation in lip reading, motivated by two observations. Firstly, a speaker's own characteristics can always be portrayed well by his/her few facial images or even a single image with shallow networks, while the fine-grained dynamic features associated with speech content expressed by the talking face always need deep sequential networks to represent accurately. Therefore, we treat the shallow and deep layers differently for speaker adaptive lip reading. Secondly, we observe that a speaker's unique characteristics ( e.g. prominent oral cavity and mandible) have varied effects on lip reading performance for different words and pronunciations, necessitating adaptive enhancement or suppression of the features for robust lip reading. Based on these two observations, we propose to take advantage of the speaker's own characteristics to automatically learn separable hidden unit contributions with different targets for shallow layers and deep layers respectively. For shallow layers where features related to the speaker's characteristics are stronger than the speech content related features, we introduce speaker-adaptive features to learn for enhancing the speech content features. For deep layers where both the speaker's features and the speech content features are all expressed well, we introduce the speaker-adaptive features to learn for suppressing the speech content irrelevant noise for robust lip reading. Our approach consistently outperforms existing methods, as confirmed by comprehensive analysis and comparison across different settings. Besides the evaluation on the popular LRW-ID and GRID datasets, we also release a new dataset for evaluation, CAS-VSR-S68h, to further assess the performance in an extreme setting where just a few speakers are available but the speech content covers a large and diversified range.

Feixiang Wang, Shuang Yang, Shiguang Shan et al.

In this work, we focus on leveraging facial cues beyond the lip region for robust Audio-Visual Speech Enhancement (AVSE). The facial region, encompassing the lip region, reflects additional speech-related attributes such as gender, skin color, nationality, etc., which contribute to the effectiveness of AVSE. However, static and dynamic speech-unrelated attributes also exist, causing appearance changes during speech. To address these challenges, we propose a Dual Attention Cooperative Framework, DualAVSE, to ignore speech-unrelated information, capture speech-related information with facial cues, and dynamically integrate it with the audio signal for AVSE. Specifically, we introduce a spatial attention-based visual encoder to capture and enhance visual speech information beyond the lip region, incorporating global facial context and automatically ignoring speech-unrelated information for robust visual feature extraction. Additionally, a dynamic visual feature fusion strategy is introduced by integrating a temporal-dimensional self-attention module, enabling the model to robustly handle facial variations. The acoustic noise in the speaking process is variable, impacting audio quality. Therefore, a dynamic fusion strategy for both audio and visual features is introduced to address this issue. By integrating cooperative dual attention in the visual encoder and audio-visual fusion strategy, our model effectively extracts beneficial speech information from both audio and visual cues for AVSE. Thorough analysis and comparison on different datasets, including normal and challenging cases with unreliable or absent visual information, consistently show our model outperforming existing methods across multiple metrics.

Shaolei Zhang, Qingkai Fang, Zhuocheng Zhang et al.

Large language models (LLMs) have demonstrated remarkable prowess in language understanding and generation. Advancing from foundation LLMs to instructionfollowing LLMs, instruction tuning plays a vital role in aligning LLMs to human preferences. However, the existing LLMs are usually focused on English, leading to inferior performance in non-English languages. In order to improve the performance for non-English languages, it is necessary to collect language-specific training data for foundation LLMs and construct language-specific instructions for instruction tuning, both of which are heavy loads. To minimize human workload, we propose to transfer the capabilities of language generation and instruction following from English to other languages through an interactive translation task. We have developed BayLing, an instruction-following LLM by utilizing LLaMA as the foundation LLM and automatically constructing interactive translation instructions for instructing tuning. Extensive assessments demonstrate that BayLing achieves comparable performance to GPT-3.5-turbo, despite utilizing a considerably smaller parameter size of only 13 billion. Experimental results on translation tasks show that BayLing achieves 95% of single-turn translation capability compared to GPT-4 with automatic evaluation and 96% of interactive translation capability compared to GPT-3.5-turbo with human evaluation. To estimate the performance on general tasks, we created a multi-turn instruction test set called BayLing-80. The experimental results on BayLing-80 indicate that BayLing achieves 89% of performance compared to GPT-3.5-turbo. BayLing also demonstrates outstanding performance on knowledge assessment of Chinese GaoKao and English SAT, second only to GPT-3.5-turbo among a multitude of instruction-following LLMs. Demo, homepage, code and models of BayLing are available.

Junqi Yang, Yuecong Min, Jie Zhang et al.

Despite rapid progress, Video Large Language Models (Video-LLMs) remain unreliable due to hallucinations, which are outputs that contradict either video evidence (faithfulness) or verifiable world knowledge (factuality). Existing benchmarks provide limited coverage of factuality hallucinations and predominantly evaluate models only in clean settings. We introduce \textsc{INFACT}, a diagnostic benchmark comprising 9{,}800 QA instances with fine-grained taxonomies for faithfulness and factuality, spanning real and synthetic videos. \textsc{INFACT} evaluates models in four modes: Base (clean), Visual Degradation, Evidence Corruption, and Temporal Intervention for order-sensitive items. Reliability under induced modes is quantified using Resist Rate (RR) and Temporal Sensitivity Score (TSS). Experiments on 14 representative Video-LLMs reveal that higher Base-mode accuracy does not reliably translate to higher reliability in the induced modes, with evidence corruption reducing stability and temporal intervention yielding the largest degradation. Notably, many open-source baselines exhibit near-zero TSS on factuality, indicating pronounced temporal inertia on order-sensitive questions.

Bin Fu, Qiyang Wan, Jialin Li et al.

Categorization, a core cognitive ability in humans that organizes objects based on common features, is essential to cognitive science as well as computer vision. To evaluate the categorization ability of visual AI models, various proxy tasks on recognition from datasets to open world scenarios have been proposed. Recent development of Large Multimodal Models (LMMs) has demonstrated impressive results in high-level visual tasks, such as visual question answering, video temporal reasoning, etc., utilizing the advanced architectures and large-scale multimodal instruction tuning. Previous researchers have developed holistic benchmarks to measure the high-level visual capability of LMMs, but there is still a lack of pure and in-depth quantitative evaluation of the most fundamental categorization ability. According to the research on human cognitive process, categorization can be seen as including two parts: category learning and category use. Inspired by this, we propose a novel, challenging, and efficient benchmark based on composite blocks, called ComBo, which provides a disentangled evaluation framework and covers the entire categorization process from learning to use. By analyzing the results of multiple evaluation tasks, we find that although LMMs exhibit acceptable generalization ability in learning new categories, there are still gaps compared to humans in many ways, such as fine-grained perception of spatial relationship and abstract category understanding. Through the study of categorization, we can provide inspiration for the further development of LMMs in terms of interpretability and generalization.

Senwei Xie, Yuntian Zhang, Ruiping Wang et al.

While skill-centric approaches leverage foundation models to enhance generalization in compositional tasks, they often rely on fixed skill libraries, limiting adaptability to new tasks without manual intervention. To address this, we propose Uni-Skill, a Unified Skill-centric framework that supports skill-aware planning and facilitates automatic skill evolution. Unlike prior methods that restrict planning to predefined skills, Uni-Skill requests for new skill implementations when existing ones are insufficient, ensuring adaptable planning with self-augmented skill library. To support automatic implementation of diverse skills requested by the planning module, we construct SkillFolder, a VerbNet-inspired repository derived from large-scale unstructured robotic videos. SkillFolder introduces a hierarchical skill taxonomy that captures diverse skill descriptions at multiple levels of abstraction. By populating this taxonomy with large-scale, automatically annotated demonstrations, Uni-Skill shifts the paradigm of skill acquisition from inefficient manual annotation to efficient offline structural retrieval. Retrieved examples provide semantic supervision over behavior patterns and fine-grained references for spatial trajectories, enabling few-shot skill inference without deployment-time demonstrations. Comprehensive experiments in both simulation and real-world settings verify the state-of-the-art performance of Uni-Skill over existing VLM-based skill-centric approaches, highlighting its advanced reasoning capabilities and strong zero-shot generalization across a wide range of novel tasks.

Ziwei Yao, Qiyang Wan, Ruiping Wang et al.

Understanding objects is fundamental to computer vision. Beyond object recognition that provides only a category label as typical output, in-depth object understanding represents a comprehensive perception of an object category, involving its components, appearance characteristics, inter-category relationships, contextual background knowledge, etc. Developing such capability requires sufficient multi-modal data, including visual annotations such as parts, attributes, and co-occurrences for specific tasks, as well as textual knowledge to support high-level tasks like reasoning and question answering. However, these data are generally task-oriented and not systematically organized enough to achieve the expected understanding of object categories. In response, we propose the Visual Knowledge Base that structures multi-modal object knowledge as graphs, and present a construction framework named VisKnow that extracts multi-modal, object-level knowledge for object understanding. This framework integrates enriched aligned text and image-source knowledge with region annotations at both object and part levels through a combination of expert design and large-scale model application. As a specific case study, we construct AnimalKB, a structured animal knowledge base covering 406 animal categories, which contains 22K textual knowledge triplets extracted from encyclopedic documents, 420K images, and corresponding region annotations. A series of experiments showcase how AnimalKB enhances object-level visual tasks such as zero-shot recognition and fine-grained VQA, and serves as challenging benchmarks for knowledge graph completion and part segmentation. Our findings highlight the potential of automatically constructing visual knowledge bases to advance visual understanding and its practical applications. The project page is available at https://vipl-vsu.github.io/VisKnow.

Wenrui Liu, Hong Chang, Ruibing Hou et al.

Out-of-Distribution (OOD) detection requires sensitivity to subtle shifts without overreacting to natural In-Distribution (ID) diversity. However, from the viewpoint of detection granularity, global representation inevitably suppress local OOD cues, while patch-based methods are unstable due to entangled spurious-correlation and noise. And neither them is effective in detecting compositional OODs composed of valid ID components. Inspired by recognition-by-components theory, we present a training-free Component-Based OOD Detection (CoOD) framework that addresses the existing limitations by decomposing inputs into functional components. To instantiate CoOD, we derive Component Shift Score (CSS) to detect local appearance shifts, and Compositional Consistency Score (CCS) to identify cross-component compositional inconsistencies. Empirically, CoOD achieves consistent improvements on both coarse- and fine-grained OOD detection.

Xiaolong Tang, Meina Kan, Shiguang Shan et al.

Predicting the trajectories of road agents is essential for autonomous driving systems. The recent mainstream methods follow a static paradigm, which predicts the future trajectory by using a fixed duration of historical frames. These methods make the predictions independently even at adjacent time steps, which leads to potential instability and temporal inconsistency. As successive time steps have largely overlapping historical frames, their forecasting should have intrinsic correlation, such as overlapping predicted trajectories should be consistent, or be different but share the same motion goal depending on the road situation. Motivated by this, in this work, we introduce HPNet, a novel dynamic trajectory forecasting method. Aiming for stable and accurate trajectory forecasting, our method leverages not only historical frames including maps and agent states, but also historical predictions. Specifically, we newly design a Historical Prediction Attention module to automatically encode the dynamic relationship between successive predictions. Besides, it also extends the attention range beyond the currently visible window benefitting from the use of historical predictions. The proposed Historical Prediction Attention together with the Agent Attention and Mode Attention is further formulated as the Triple Factorized Attention module, serving as the core design of HPNet.Experiments on the Argoverse and INTERACTION datasets show that HPNet achieves state-of-the-art performance, and generates accurate and stable future trajectories. Our code are available at https://github.com/XiaolongTang23/HPNet.

Ruibing Hou, Mingyue Zhou, Yuwei Gui et al.

Faithfully modeling human behavior in dynamic environments is a foundational challenge for embodied intelligence. While conditional motion synthesis has achieved significant advances, egocentric motion generation remains largely underexplored due to the inherent complexity of first-person perception. In this work, we investigate Egocentric Vision-Language (Ego-VL) motion generation. This task requires synthesizing 3D human motion conditioned jointly on first-person visual observations and natural language instructions. We identify a critical \textit{reasoning-generation entanglement} challenge: the simultaneous optimization of semantic reasoning and kinematic modeling introduces gradient conflicts. These conflicts systematically degrade the fidelity of multimodal grounding and motion quality. To address this challenge, we propose a hierarchical generative framework \textbf{EgoMotion}. Inspired by the biological decoupling of cognitive reasoning and motor control, EgoMotion operates in two stages. In the Cognitive Reasoning stage, A vision-language model (VLM) projects multimodal inputs into a structured space of discrete motion primitives. This forces the VLM to acquire goal-consistent representations, effectively bridging the semantic gap between high-level perceptual understanding and low-level action execution. In the Motion Generation stage, these learned representations serve as expressive conditioning signals for a diffusion-based motion generator. By performing iterative denoising within a continuous latent space, the generator synthesizes physically plausible and temporally coherent trajectories. Extensive evaluations demonstrate that EgoMotion achieves state-of-the-art performance, and produces motion sequences that are both semantically grounded and kinematically superior to existing approaches.

Jie Zhang, Shuai Dong, Shiguang Shan et al.

Recent progress in text-to-image diffusion models has transformed image editing via text prompts, yet this also introduces significant ethical challenges from potential misuse in creating deceptive or harmful content. While current defenses seek to mitigate this risk by embedding imperceptible perturbations, their effectiveness is limited against malicious tampering. To address this issue, we propose a Dual Attention-Guided Noise Perturbation (DANP) immunization method that adds imperceptible perturbations to disrupt the model's semantic understanding and generation process. DANP functions over multiple timesteps to manipulate both cross-attention maps and the noise prediction process, using a dynamic threshold to generate masks that identify text-relevant and irrelevant regions. It then reduces attention in relevant areas while increasing it in irrelevant ones, thereby misguides the edit towards incorrect regions and preserves the intended targets. Additionally, our method maximizes the discrepancy between the injected noise and the model's predicted noise to further interfere with the generation. By targeting both attention and noise prediction mechanisms, DANP exhibits impressive immunity against malicious edits, and extensive experiments confirm that our method achieves state-of-the-art performance.

Shuai Dong, Jie Zhang, Guoying Zhao et al.

Text-guided image editing via diffusion models, while powerful, raises significant concerns about misuse, motivating efforts to immunize images against unauthorized edits using imperceptible perturbations. Prevailing metrics for evaluating immunization success typically rely on measuring the visual dissimilarity between the output generated from a protected image and a reference output generated from the unprotected original. This approach fundamentally overlooks the core requirement of image immunization, which is to disrupt semantic alignment with attacker intent, regardless of deviation from any specific output. We argue that immunization success should instead be defined by the edited output either semantically mismatching the prompt or suffering substantial perceptual degradations, both of which thwart malicious intent. To operationalize this principle, we propose Synergistic Intermediate Feature Manipulation (SIFM), a method that strategically perturbs intermediate diffusion features through dual synergistic objectives: (1) maximizing feature divergence from the original edit trajectory to disrupt semantic alignment with the expected edit, and (2) minimizing feature norms to induce perceptual degradations. Furthermore, we introduce the Immunization Success Rate (ISR), a novel metric designed to rigorously quantify true immunization efficacy for the first time. ISR quantifies the proportion of edits where immunization induces either semantic failure relative to the prompt or significant perceptual degradations, assessed via Multimodal Large Language Models (MLLMs). Extensive experiments show our SIFM achieves the state-of-the-art performance for safeguarding visual content against malicious diffusion-based manipulation.

Ziyi Bai, Ruiping Wang, Xilin Chen

Video Question Answering (VideoQA) has emerged as a vital tool to evaluate agents' ability to understand human daily behaviors. Despite the recent success of large vision language models in many multi-modal tasks, complex situation reasoning over videos involving multiple human-object interaction events still remains challenging. In contrast, humans can easily tackle it by using a series of episode memories as anchors to quickly locate question-related key moments for reasoning. To mimic this effective reasoning strategy, we propose the Glance-Focus model. One simple way is to apply an action detection model to predict a set of actions as key memories. However, these actions within a closed set vocabulary are hard to generalize to various video domains. Instead of that, we train an Encoder-Decoder to generate a set of dynamic event memories at the glancing stage. Apart from using supervised bipartite matching to obtain the event memories, we further design an unsupervised memory generation method to get rid of dependence on event annotations. Next, at the focusing stage, these event memories act as a bridge to establish the correlation between the questions with high-level event concepts and low-level lengthy video content. Given the question, the model first focuses on the generated key event memory, then focuses on the most relevant moment for reasoning through our designed multi-level cross-attention mechanism. We conduct extensive experiments on four Multi-Event VideoQA benchmarks including STAR, EgoTaskQA, AGQA, and NExT-QA. Our proposed model achieves state-of-the-art results, surpassing current large models in various challenging reasoning tasks. The code and models are available at https://github.com/ByZ0e/Glance-Focus.

Sen Nie, Jie Zhang, Zhuo Wang et al.

Vision-language models (VLMs) such as CLIP have demonstrated remarkable zero-shot generalization, yet remain highly vulnerable to adversarial examples (AEs). While test-time defenses are promising, existing methods fail to provide sufficient robustness against strong attacks and are often hampered by high inference latency and task-specific applicability. To address these limitations, we start by investigating the intrinsic properties of AEs, which reveals that AEs exhibit severe feature inconsistency under progressive frequency attenuation. We further attribute this to the model's inherent spectral bias. Leveraging this insight, we propose an efficient test-time defense named Contrastive Spectral Rectification (CSR). CSR optimizes a rectification perturbation to realign the input with the natural manifold under a spectral-guided contrastive objective, which is applied input-adaptively. Extensive experiments across 16 classification benchmarks demonstrate that CSR outperforms the SOTA by an average of 18.1% against strong AutoAttack with modest inference overhead. Furthermore, CSR exhibits broad applicability across diverse visual tasks. Code is available at https://github.com/Summu77/CSR.

Hongyu Wang, Chuyan Xiong, Ruiping Wang et al.

Vision-Language-Action (VLA) models have shown impressive capabilities across a wide range of robotics manipulation tasks. However, their growing model size poses significant challenges for deployment on resource-constrained robotic systems. While 1-bit pretraining has proven effective for enhancing the inference efficiency of large language models with minimal performance loss, its application to VLA models remains underexplored. In this work, we present BitVLA, the first 1-bit VLA model for robotics manipulation, in which every parameter is ternary, i.e., {-1, 0, 1}. To further reduce the memory footprint of the vision encoder, we propose the distillation-aware training strategy that compresses the full-precision encoder to 1.58-bit weights. During this process, a full-precision encoder serves as a teacher model to better align latent representations. Despite the lack of large-scale robotics pretraining, BitVLA achieves performance comparable to the state-of-the-art model OpenVLA-OFT with 4-bit post-training quantization on the LIBERO benchmark, while consuming only 29.8% of the memory. These results highlight BitVLA's promise for deployment on memory-constrained edge devices. We release the code and model weights in https://github.com/ustcwhy/BitVLA.

Yifeng Xu, Zhenliang He, Shiguang Shan et al.

Recently, large-scale diffusion models have made impressive progress in text-to-image (T2I) generation. To further equip these T2I models with fine-grained spatial control, approaches like ControlNet introduce an extra network that learns to follow a condition image. However, for every single condition type, ControlNet requires independent training on millions of data pairs with hundreds of GPU hours, which is quite expensive and makes it challenging for ordinary users to explore and develop new types of conditions. To address this problem, we propose the CtrLoRA framework, which trains a Base ControlNet to learn the common knowledge of image-to-image generation from multiple base conditions, along with condition-specific LoRAs to capture distinct characteristics of each condition. Utilizing our pretrained Base ControlNet, users can easily adapt it to new conditions, requiring as few as 1,000 data pairs and less than one hour of single-GPU training to obtain satisfactory results in most scenarios. Moreover, our CtrLoRA reduces the learnable parameters by 90% compared to ControlNet, significantly lowering the threshold to distribute and deploy the model weights. Extensive experiments on various types of conditions demonstrate the efficiency and effectiveness of our method. Codes and model weights will be released at https://github.com/xyfJASON/ctrlora.

Bei Yan, Jie Zhang, Zhiyuan Chen et al.

Recently, large foundation models, including large language models (LLMs) and large vision-language models (LVLMs), have become essential tools in critical fields such as law, finance, and healthcare. As these models increasingly integrate into our daily life, it is necessary to conduct moral evaluation to ensure that their outputs align with human values and remain within moral boundaries. Previous works primarily focus on LLMs, proposing moral datasets and benchmarks limited to text modality. However, given the rapid development of LVLMs, there is still a lack of multimodal moral evaluation methods. To bridge this gap, we introduce M$^3$oralBench, the first MultiModal Moral Benchmark for LVLMs. M$^3$oralBench expands the everyday moral scenarios in Moral Foundations Vignettes (MFVs) and employs the text-to-image diffusion model, SD3.0, to create corresponding scenario images. It conducts moral evaluation across six moral foundations of Moral Foundations Theory (MFT) and encompasses tasks in moral judgement, moral classification, and moral response, providing a comprehensive assessment of model performance in multimodal moral understanding and reasoning. Extensive experiments on 10 popular open-source and closed-source LVLMs demonstrate that M$^3$oralBench is a challenging benchmark, exposing notable moral limitations in current models. Our benchmark is publicly available.

Changzhen Li, Yuecong Min, Jie Zhang et al.

The rapid evolution of Text-to-Video (T2V) diffusion models has driven remarkable advancements in generating high-quality, temporally coherent videos from natural language descriptions. Despite these achievements, their vulnerability to adversarial attacks remains largely unexplored. In this paper, we introduce T2VAttack, a comprehensive study of adversarial attacks on T2V diffusion models from both semantic and temporal perspectives. Considering the inherently dynamic nature of video data, we propose two distinct attack objectives: a semantic objective to evaluate video-text alignment and a temporal objective to assess the temporal dynamics. To achieve an effective and efficient attack process, we propose two adversarial attack methods: (i) T2VAttack-S, which identifies semantically or temporally critical words in prompts and replaces them with synonyms via greedy search, and (ii) T2VAttack-I, which iteratively inserts optimized words with minimal perturbation to the prompt. By combining these objectives and strategies, we conduct a comprehensive evaluation on the adversarial robustness of several state-of-the-art T2V models, including ModelScope, CogVideoX, Open-Sora, and HunyuanVideo. Our experiments reveal that even minor prompt modifications, such as the substitution or insertion of a single word, can cause substantial degradation in semantic fidelity and temporal dynamics, highlighting critical vulnerabilities in current T2V diffusion models.

Xuanyu Ge, Zhongqi Wang, Jie Zhang et al.

Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities across various tasks, yet they remain vulnerable to backdoor attacks. Existing defense methods predominantly focus on sample-level defense, which relies on the knowledge of training data or triggers. However, identifying whether a given model is backdoored remains a critical but unexplored task. To fill this gap, we propose EntropyScan, a lightweight and trigger-agnostic method for model-level backdoor detection in LVLMs. We first observe that backdoor injection disrupts the cross-modal alignment, resulting in pronounced structural anomalies in visual attention allocation on benign samples. Based on this insight, EntropyScan detects the backdoor models by quantifying such attention deviations. Specifically, it extracts visual attention distributions from the initial layers of the Large Language Model (LLM) and applies Tsallis entropy to capture these structural distortions. By employing a reference-anchored Z-score normalization on a small set of benign samples, it effectively identifies the backdoored model. Extensive experiments across two LVLMs architectures and three advanced attack scenarios show that EntropyScan achieves an F1 score of 98.5% in average and an AUC of 96.6%. Our code will be publicly available soon.

Xiaolong Tang, Meina Kan, Shiguang Shan et al.

Safe and feasible trajectory planning is critical for real-world autonomous driving systems. However, existing learning-based planners rely heavily on expert demonstrations, which not only lack explicit safety awareness but also risk inheriting undesirable behaviors such as speeding from suboptimal human driving data. Inspired by the success of large language models, we propose Plan-R1, a two-stage trajectory planning framework that decouples principle alignment from behavior learning. In the first stage, a general trajectory predictor is pre-trained on expert data to capture diverse, human-like driving behaviors. In the second stage, the model is fine-tuned with rule-based rewards using Group Relative Policy Optimization (GRPO), explicitly aligning ego planning with principles such as safety, comfort, and traffic rule compliance. This two-stage paradigm retains human-like behaviors while enhancing safety awareness and discarding undesirable patterns from demonstrations. Furthermore, we identify a key limitation of directly applying GRPO to planning: group-wise normalization erases cross-group scale differences, causing rare, high-variance safety-violation groups to have similar advantages as abundant low-variance safe groups, thereby suppressing optimization for safety-critical objectives. To address this, we propose Variance-Decoupled GRPO (VD-GRPO), which replaces normalization with centering and fixed scaling to preserve absolute reward magnitudes, ensuring that safety-critical objectives remain dominant throughout training. Experiments on the nuPlan benchmark demonstrate that Plan-R1 significantly improves planning safety and feasibility, achieving state-of-the-art performance, particularly in realistic reactive settings. Our code is available at https://github.com/XiaolongTang23/Plan-R1.

Wenzhao Xiang, Yue Wu, Hongyang Yu et al.

Self-supervised visual pre-training methods face an inherent tension: contrastive learning (CL) captures global semantics but loses fine-grained detail, while masked image modeling (MIM) preserves local textures but suffers from "attention drift" due to semantically-agnostic random masking. We propose C2FMAE, a coarse-to-fine masked autoencoder that resolves this tension by explicitly learning hierarchical visual representations across three data granularities: semantic masks (scene-level), instance masks (object-level), and RGB images (pixel-level). Two synergistic innovations enforce a strict top-down learning principle. First, a cascaded decoder sequentially reconstructs from scene semantics to object instances to pixel details, establishing explicit cross-granularity dependencies that parallel decoders cannot capture. Second, a progressive masking curriculum dynamically shifts the training focus from semantic-guided to instance-guided and finally to random masking, creating a structured learning path from global context to local features. To support this framework, we construct a large-scale multi-granular dataset with high-quality pseudo-labels for all 1.28M ImageNet-1K images. Extensive experiments show that C2FMAE achieves significant performance gains on image classification, object detection, and semantic segmentation, validating the effectiveness of our hierarchical design in learning more robust and generalizable representations.

Yinqi Li, Jiahe Zhao, Hong Chang et al.

Contrastive Language-Image Pre-training (CLIP) has become a foundation model and has been applied to various vision and multimodal tasks. However, recent works indicate that CLIP falls short in distinguishing detailed differences in images and shows suboptimal performance on dense-prediction and vision-centric multimodal tasks. Therefore, this work focuses on improving existing CLIP models, aiming to capture as many visual details in images as possible. We find that a specific type of generative models, unCLIP, provides a suitable framework for achieving our goal. Specifically, unCLIP trains an image generator conditioned on the CLIP image embedding. In other words, it inverts the CLIP image encoder. Compared to discriminative models like CLIP, generative models are better at capturing image details because they are trained to learn the data distribution of images. Additionally, the conditional input space of unCLIP aligns with CLIP's original image-text embedding space. Therefore, we propose to invert unCLIP (dubbed un$^2$CLIP) to improve the CLIP model. In this way, the improved image encoder can gain unCLIP's visual detail capturing ability while preserving its alignment with the original text encoder simultaneously. We evaluate our improved CLIP across various tasks to which CLIP has been applied, including the challenging MMVP-VLM benchmark, the dense-prediction open-vocabulary segmentation task, and multimodal large language model tasks. Experiments show that un$^2$CLIP significantly improves the original CLIP and previous CLIP improvement methods. Code and models will be available at https://github.com/LiYinqi/un2CLIP.

Jie Zhang, Zheng Yuan, Zhongqi Wang et al.

The rapid evolution of Large Vision-Language Models (LVLMs) has highlighted the necessity for comprehensive evaluation frameworks that assess these models across diverse dimensions. While existing benchmarks focus on specific aspects such as perceptual abilities, cognitive capabilities, and safety against adversarial attacks, they often lack the breadth and depth required to provide a holistic understanding of LVLMs' strengths and limitations. To address this gap, we introduce REVAL, a comprehensive benchmark designed to evaluate the \textbf{RE}liability and \textbf{VAL}ue of LVLMs. REVAL encompasses over 144K image-text Visual Question Answering (VQA) samples, structured into two primary sections: Reliability, which assesses truthfulness (\eg, perceptual accuracy and hallucination tendencies) and robustness (\eg, resilience to adversarial attacks, typographic attacks, and image corruption), and Values, which evaluates ethical concerns (\eg, bias and moral understanding), safety issues (\eg, toxicity and jailbreak vulnerabilities), and privacy problems (\eg, privacy awareness and privacy leakage). We evaluate 26 models, including mainstream open-source LVLMs and prominent closed-source models like GPT-4o and Gemini-1.5-Pro. Our findings reveal that while current LVLMs excel in perceptual tasks and toxicity avoidance, they exhibit significant vulnerabilities in adversarial scenarios, privacy preservation, and ethical reasoning. These insights underscore critical areas for future improvements, guiding the development of more secure, reliable, and ethically aligned LVLMs. REVAL provides a robust framework for researchers to systematically assess and compare LVLMs, fostering advancements in the field.

Zhongqi Wang, Jie Zhang, Shiguang Shan et al.

Recent studies have revealed that text-to-image diffusion models are vulnerable to backdoor attacks, where attackers implant stealthy textual triggers to manipulate model outputs. Previous backdoor detection methods primarily focus on the static features of backdoor samples. However, a vital property of diffusion models is their inherent dynamism. This study introduces a novel backdoor detection perspective named Dynamic Attention Analysis (DAA), showing that these dynamic characteristics serve as better indicators for backdoor detection. Specifically, by examining the dynamic evolution of cross-attention maps, we observe that backdoor samples exhibit distinct feature evolution patterns at the $<$EOS$>$ token compared to benign samples. To quantify these dynamic anomalies, we first introduce DAA-I, which treats the tokens' attention maps as spatially independent and measures dynamic feature using the Frobenius norm. Furthermore, to better capture the interactions between attention maps and refine the feature, we propose a dynamical system-based approach, referred to as DAA-S. This model formulates the spatial correlations among attention maps using a graph-based state equation and we theoretically analyze the global asymptotic stability of this method. Extensive experiments across five representative backdoor attack scenarios demonstrate that our approach significantly surpasses existing detection methods, achieving an average F1 Score of 79.49% and an AUC of 87.67%. The code is available at https://github.com/Robin-WZQ/DAA.

Jiachen Liang, Ruibing Hou, Minyang Hu et al.

Pre-trained vision-language models (e.g., CLIP) have shown powerful zero-shot transfer capabilities. But they still struggle with domain shifts and typically require labeled data to adapt to downstream tasks, which could be costly. In this work, we aim to leverage unlabeled data that naturally spans multiple domains to enhance the transferability of vision-language models. Under this unsupervised multi-domain setting, we have identified inherent model bias within CLIP, notably in its visual and text encoders. Specifically, we observe that CLIP's visual encoder tends to prioritize encoding domain over discriminative category information, meanwhile its text encoder exhibits a preference for domain-relevant classes. To mitigate this model bias, we propose a training-free and label-free feature calibration method, Unsupervised Multi-domain Feature Calibration (UMFC). UMFC estimates image-level biases from domain-specific features and text-level biases from the direction of domain transition. These biases are subsequently subtracted from original image and text features separately, to render them domain-invariant. We evaluate our method on multiple settings including transductive learning and test-time adaptation. Extensive experiments show that our method outperforms CLIP and performs on par with the state-of-the-arts that need additional annotations or optimization. Our code is available at https://github.com/GIT-LJc/UMFC.

Minyang Hu, Hong Chang, Zong Guo et al.

Few-shot learning (FSL) aims to learn novel tasks with very few labeled samples by leveraging experience from \emph{related} training tasks. In this paper, we try to understand FSL by delving into two key questions: (1) How to quantify the relationship between \emph{training} and \emph{novel} tasks? (2) How does the relationship affect the \emph{adaptation difficulty} on novel tasks for different models? To answer the two questions, we introduce Task Attribute Distance (TAD) built upon attributes as a metric to quantify the task relatedness. Unlike many existing metrics, TAD is model-agnostic, making it applicable to different FSL models. Then, we utilize TAD metric to establish a theoretical connection between task relatedness and task adaptation difficulty. By deriving the generalization error bound on a novel task, we discover how TAD measures the adaptation difficulty on novel tasks for FSL models. To validate our TAD metric and theoretical findings, we conduct experiments on three benchmarks. Our experimental results confirm that TAD metric effectively quantifies the task relatedness and reflects the adaptation difficulty on novel tasks for various FSL methods, even if some of them do not learn attributes explicitly or human-annotated attributes are not available. Finally, we present two applications of the proposed TAD metric: data augmentation and test-time intervention, which further verify its effectiveness and general applicability. The source code is available at https://github.com/hu-my/TaskAttributeDistance.

Sen Nie, Jie Zhang, Jianxin Yan et al.

Adversarial attacks have evolved from simply disrupting predictions on conventional task-specific models to the more complex goal of manipulating image semantics on Large Vision-Language Models (LVLMs). However, existing methods struggle with controllability and fail to precisely manipulate the semantics of specific concepts in the image. We attribute this limitation to semantic entanglement in the patch-token representations on which adversarial attacks typically operate: global context aggregated by self-attention in the vision encoder dominates individual patch features, making them unreliable handles for precise local semantic manipulation. Our systematic investigation reveals a key insight: value features (V) computed within the transformer attention block serve as much more precise handles for manipulation. We show that V suppresses global-context channels, allowing it to retain high-entropy, disentangled local semantic information. Building on this discovery, we propose V-Attack, a novel method designed for precise local semantic attacks. V-Attack targets the value features and introduces two core components: (1) a Self-Value Enhancement module to refine V's intrinsic semantic richness, and (2) a Text-Guided Value Manipulation module that leverages text prompts to locate source concept and optimize it toward a target concept. By bypassing the entangled patch features, V-Attack achieves highly effective semantic control. Extensive experiments across diverse LVLMs, including LLaVA, InternVL, DeepseekVL and GPT-4o, show that V-Attack improves the attack success rate by an average of 36% over state-of-the-art methods, exposing critical vulnerabilities in modern visual-language understanding. Our code and data are available https://github.com/Summu77/V-Attack.

Jiachen Liang, Ruibing Hou, Minyang Hu et al.

Out-of-distribution (OOD) detection is critical for ensuring the reliability of deep learning models in open-world applications. While post-hoc methods are favored for their efficiency and ease of deployment, existing approaches often underexploit the rich information embedded in the model's logits space. In this paper, we propose LogitGap, a novel post-hoc OOD detection method that explicitly exploits the relationship between the maximum logit and the remaining logits to enhance the separability between in-distribution (ID) and OOD samples. To further improve its effectiveness, we refine LogitGap by focusing on a more compact and informative subset of the logit space. Specifically, we introduce a training-free strategy that automatically identifies the most informative logits for scoring. We provide both theoretical analysis and empirical evidence to validate the effectiveness of our approach. Extensive experiments on both vision-language and vision-only models demonstrate that LogitGap consistently achieves state-of-the-art performance across diverse OOD detection scenarios and benchmarks. Code is available at https://github.com/GIT-LJc/LogitGap.

Zaifei Yang, Hong Chang, Ruibing Hou et al.

The molecular large language models have garnered widespread attention due to their promising potential on molecular applications. However, current molecular large language models face significant limitations in understanding molecules due to inadequate textual descriptions and suboptimal molecular representation strategies during pretraining. To address these challenges, we introduce KnowMol-100K, a large-scale dataset with 100K fine-grained molecular annotations across multiple levels, bridging the gap between molecules and textual descriptions. Additionally, we propose chemically-informative molecular representation, effectively addressing limitations in existing molecular representation strategies. Building upon these innovations, we develop KnowMol, a state-of-the-art multi-modal molecular large language model. Extensive experiments demonstrate that KnowMol achieves superior performance across molecular understanding and generation tasks. GitHub: https://github.com/yzf-code/KnowMol Huggingface: https://hf.co/datasets/yzf1102/KnowMol-100K

Yifeng Xu, Zhenliang He, Meina Kan et al.

Visual generation and understanding are two deeply interconnected aspects of human intelligence, yet they have been traditionally treated as separate tasks in machine learning. In this paper, we propose Jodi, a diffusion framework that unifies visual generation and understanding by jointly modeling the image domain and multiple label domains. Specifically, Jodi is built upon a linear diffusion transformer along with a role switch mechanism, which enables it to perform three particular types of tasks: (1) joint generation, where the model simultaneously generates images and multiple labels; (2) controllable generation, where images are generated conditioned on any combination of labels; and (3) image perception, where multiple labels can be predicted at once from a given image. Furthermore, we present the Joint-1.6M dataset, which contains 200,000 high-quality images collected from public sources, automatic labels for 7 visual domains, and LLM-generated captions. Extensive experiments demonstrate that Jodi excels in both generation and understanding tasks and exhibits strong extensibility to a wider range of visual domains. Code is available at https://github.com/VIPL-GENUN/Jodi.

Yinqi Li, Hong Chang, Ruibing Hou et al.

Diffusion models have shown remarkable progress in various generative tasks such as image and video generation. This paper studies the problem of leveraging pretrained diffusion models for performing discriminative tasks. Specifically, we extend the discriminative capability of pretrained frozen generative diffusion models from the classification task to the more complex object detection task, by "inverting" a pretrained layout-to-image diffusion model. To this end, a gradient-based discrete optimization approach for replacing the heavy prediction enumeration process, and a prior distribution model for making more accurate use of the Bayes' rule, are proposed respectively. Empirical results show that this method is on par with basic discriminative object detection baselines on COCO dataset. In addition, our method can greatly speed up the previous diffusion-based method for classification without sacrificing accuracy. Code and models are available at https://github.com/LiYinqi/DIVE .

Yufei Cai, Hu Han, Yuxiang Wei et al.

The progress on generative models has led to significant advances on text-to-video (T2V) generation, yet the motion controllability of generated videos remains limited. Existing motion transfer methods explored the motion representations of reference videos to guide generation. Nevertheless, these methods typically rely on sample-specific optimization strategy, resulting in high computational burdens. In this paper, we propose EfficientMT, a novel and efficient end-to-end framework for video motion transfer. By leveraging a small set of synthetic paired motion transfer samples, EfficientMT effectively adapts a pretrained T2V model into a general motion transfer framework that can accurately capture and reproduce diverse motion patterns. Specifically, we repurpose the backbone of the T2V model to extract temporal information from reference videos, and further propose a scaler module to distill motion-related information. Subsequently, we introduce a temporal integration mechanism that seamlessly incorporates reference motion features into the video generation process. After training on our self-collected synthetic paired samples, EfficientMT enables general video motion transfer without requiring test-time optimization. Extensive experiments demonstrate that our EfficientMT outperforms existing methods in efficiency while maintaining flexible motion controllability. Our code will be available https://github.com/PrototypeNx/EfficientMT.

Jie Zhang, Zhongqi Wang, Shiguang Shan et al.

Backdoor attacks targeting text-to-image diffusion models have advanced rapidly. However, current backdoor samples often exhibit two key abnormalities compared to benign samples: 1) Semantic Consistency, where backdoor prompts tend to generate images with similar semantic content even with significant textual variations to the prompts; 2) Attention Consistency, where the trigger induces consistent structural responses in the cross-attention maps. These consistencies leave detectable traces for defenders, making backdoors easier to identify. In this paper, toward stealthy backdoor samples, we propose Trigger without Trace (TwT) by explicitly mitigating these consistencies. Specifically, our approach leverages syntactic structures as backdoor triggers to amplify the sensitivity to textual variations, effectively breaking down the semantic consistency. Besides, a regularization method based on Kernel Maximum Mean Discrepancy (KMMD) is proposed to align the distribution of cross-attention responses between backdoor and benign samples, thereby disrupting attention consistency. Extensive experiments demonstrate that our method achieves a 97.5% attack success rate while exhibiting stronger resistance to defenses. It achieves an average of over 98% backdoor samples bypassing three state-of-the-art detection mechanisms, revealing the vulnerabilities of current backdoor defense methods. The code is available at https://github.com/Robin-WZQ/TwT.

Jie Zhang, Zhongqi Wang, Mengqi Lei et al.

Currently many benchmarks have been proposed to evaluate the perception ability of the Large Vision-Language Models (LVLMs). However, most benchmarks conduct questions by selecting images from existing datasets, resulting in the potential data leakage. Besides, these benchmarks merely focus on evaluating LVLMs on the realistic style images and clean scenarios, leaving the multi-stylized images and noisy scenarios unexplored. In response to these challenges, we propose a dynamic and scalable benchmark named Dysca for evaluating LVLMs by leveraging synthesis images. Specifically, we leverage Stable Diffusion and design a rule-based method to dynamically generate novel images, questions and the corresponding answers. We consider 51 kinds of image styles and evaluate the perception capability in 20 subtasks. Moreover, we conduct evaluations under 4 scenarios (i.e., Clean, Corruption, Print Attacking and Adversarial Attacking) and 3 question types (i.e., Multi-choices, True-or-false and Free-form). Thanks to the generative paradigm, Dysca serves as a scalable benchmark for easily adding new subtasks and scenarios. A total of 24 advanced open-source LVLMs and 2 close-source LVLMs are evaluated on Dysca, revealing the drawbacks of current LVLMs. The benchmark is released at https://github.com/Robin-WZQ/Dysca.

Bei Yan, Jie Zhang, Zheng Yuan et al.

Despite the outstanding performance in multimodal tasks, Large Vision-Language Models (LVLMs) have been plagued by the issue of hallucination, i.e., generating content that is inconsistent with the corresponding visual inputs. While previous works have proposed various benchmarks to evaluate this issue, the quality of these evaluations remains unverified. We observe that some of these benchmarks may produce inconsistent evaluation results across repeated tests or fail to align with human evaluation. To address this, we propose a Hallucination benchmark Quality Measurement framework (HQM), which leverages specific indicators to assess both reliability and validity. Our empirical analysis using HQM reveals and pinpoints potential evaluation issues in existing benchmarks, exposing a critical gap in current hallucination evaluation. To bridge this gap, we propose HQH, a High-Quality Hallucination benchmark, which demonstrates superior reliability and validity under HQM, serving as a credible evaluation tool. Our large-scale evaluation of popular LVLMs on HQH reveals severe hallucination problems, which occur not only in the models' main answer to a question but also in additional analysis. This highlights the necessity for future model improvements to effectively mitigate hallucinations and reduce the associated security risks in real-world applications. Our benchmark is publicly available at https://github.com/HQHBench/HQHBench.

Sibo Wang, Xiangkui Cao, Jie Zhang et al.

The emergence of Large Vision-Language Models (LVLMs) marks significant strides towards achieving general artificial intelligence. However, these advancements are accompanied by concerns about biased outputs, a challenge that has yet to be thoroughly explored. Existing benchmarks are not sufficiently comprehensive in evaluating biases due to their limited data scale, single questioning format and narrow sources of bias. To address this problem, we introduce VLBiasBench, a comprehensive benchmark designed to evaluate biases in LVLMs. VLBiasBench, features a dataset that covers nine distinct categories of social biases, including age, disability status, gender, nationality, physical appearance, race, religion, profession, social economic status, as well as two intersectional bias categories: race x gender and race x social economic status. To build a large-scale dataset, we use Stable Diffusion XL model to generate 46,848 high-quality images, which are combined with various questions to creat 128,342 samples. These questions are divided into open-ended and close-ended types, ensuring thorough consideration of bias sources and a comprehensive evaluation of LVLM biases from multiple perspectives. We conduct extensive evaluations on 15 open-source models as well as two advanced closed-source models, yielding new insights into the biases present in these models. Our benchmark is available at https://github.com/Xiangkui-Cao/VLBiasBench.

Xingming Long, Jie Zhang, Shiguang Shan et al.

Most existing out-of-distribution (OOD) detection benchmarks classify samples with novel labels as the OOD data. However, some marginal OOD samples actually have close semantic contents to the in-distribution (ID) sample, which makes determining the OOD sample a Sorites Paradox. In this paper, we construct a benchmark named Incremental Shift OOD (IS-OOD) to address the issue, in which we divide the test samples into subsets with different semantic and covariate shift degrees relative to the ID dataset. The data division is achieved through a shift measuring method based on our proposed Language Aligned Image feature Decomposition (LAID). Moreover, we construct a Synthetic Incremental Shift (Syn-IS) dataset that contains high-quality generated images with more diverse covariate contents to complement the IS-OOD benchmark. We evaluate current OOD detection methods on our benchmark and find several important insights: (1) The performance of most OOD detection methods significantly improves as the semantic shift increases; (2) Some methods like GradNorm may have different OOD detection mechanisms as they rely less on semantic shifts to make decisions; (3) Excessive covariate shifts in the image are also likely to be considered as OOD for some methods. Our code and data are released in https://github.com/qqwsad5/IS-OOD.

Jie Zhang, Xiangkui Cao, Zhouyu Han et al.

Large Vision-Language Models (LVLMs) exhibit impressive potential across various tasks but also face significant privacy risks, limiting their practical applications. Current researches on privacy assessment for LVLMs is limited in scope, with gaps in both assessment dimensions and privacy categories. To bridge this gap, we propose Multi-P$^2$A, a comprehensive benchmark for evaluating the privacy preservation capabilities of LVLMs in terms of privacy awareness and leakage. Privacy awareness measures the model's ability to recognize the privacy sensitivity of input data, while privacy leakage assesses the risk of the model unintentionally disclosing privacy information in its output. We design a range of sub-tasks to thoroughly evaluate the model's privacy protection offered by LVLMs. Multi-P$^2$A covers 26 categories of personal privacy, 15 categories of trade secrets, and 18 categories of state secrets, totaling 31,962 samples. Based on Multi-P$^2$A, we evaluate the privacy preservation capabilities of 21 open-source and 2 closed-source LVLMs. Our results reveal that current LVLMs generally pose a high risk of facilitating privacy breaches, with vulnerabilities varying across personal privacy, trade secret, and state secret.

Yuhui Yuan, Rao Fu, Lang Huang et al.

We present a High-Resolution Transformer (HRFormer) that learns high-resolution representations for dense prediction tasks, in contrast to the original Vision Transformer that produces low-resolution representations and has high memory and computational cost. We take advantage of the multi-resolution parallel design introduced in high-resolution convolutional networks (HRNet), along with local-window self-attention that performs self-attention over small non-overlapping image windows, for improving the memory and computation efficiency. In addition, we introduce a convolution into the FFN to exchange information across the disconnected image windows. We demonstrate the effectiveness of the High-Resolution Transformer on both human pose estimation and semantic segmentation tasks, e.g., HRFormer outperforms Swin transformer by $1.3$ AP on COCO pose estimation with $50\%$ fewer parameters and $30\%$ fewer FLOPs. Code is available at: https://github.com/HRNet/HRFormer.

Ruibing Hou, Bingpeng Ma, Hong Chang et al.

Person re-identification (reID) plays an important role in computer vision. However, existing methods suffer from performance degradation in occluded scenes. In this work, we propose an occlusion-robust block, Region Feature Completion (RFC), for occluded reID. Different from most previous works that discard the occluded regions, RFC block can recover the semantics of occluded regions in feature space. Firstly, a Spatial RFC (SRFC) module is developed. SRFC exploits the long-range spatial contexts from non-occluded regions to predict the features of occluded regions. The unit-wise prediction task leads to an encoder/decoder architecture, where the region-encoder models the correlation between non-occluded and occluded region, and the region-decoder utilizes the spatial correlation to recover occluded region features. Secondly, we introduce Temporal RFC (TRFC) module which captures the long-term temporal contexts to refine the prediction of SRFC. RFC block is lightweight, end-to-end trainable and can be easily plugged into existing CNNs to form RFCnet. Extensive experiments are conducted on occluded and commonly holistic reID benchmarks. Our method significantly outperforms existing methods on the occlusion datasets, while remains top even superior performance on holistic datasets. The source code is available at https://github.com/blue-blue272/OccludedReID-RFCnet.

Ruibing Hou, Bingpeng Ma, Hong Chang et al.

Person re-identification (reID) by CNNs based networks has achieved favorable performance in recent years. However, most of existing CNNs based methods do not take full advantage of spatial-temporal context modeling. In fact, the global spatial-temporal context can greatly clarify local distractions to enhance the target feature representation. To comprehensively leverage the spatial-temporal context information, in this work, we present a novel block, Interaction-Aggregation-Update (IAU), for high-performance person reID. Firstly, Spatial-Temporal IAU (STIAU) module is introduced. STIAU jointly incorporates two types of contextual interactions into a CNN framework for target feature learning. Here the spatial interactions learn to compute the contextual dependencies between different body parts of a single frame. While the temporal interactions are used to capture the contextual dependencies between the same body parts across all frames. Furthermore, a Channel IAU (CIAU) module is designed to model the semantic contextual interactions between channel features to enhance the feature representation, especially for small-scale visual cues and body parts. Therefore, the IAU block enables the feature to incorporate the globally spatial, temporal, and channel context. It is lightweight, end-to-end trainable, and can be easily plugged into existing CNNs to form IAUnet. The experiments show that IAUnet performs favorably against state-of-the-art on both image and video reID tasks and achieves compelling results on a general object categorization task. The source code is available at https://github.com/blue-blue272/ImgReID-IAnet.

Ruibing Hou, Hong Chang, Bingpeng Ma et al.

This paper proposes a Temporal Complementary Learning Network that extracts complementary features of consecutive video frames for video person re-identification. Firstly, we introduce a Temporal Saliency Erasing (TSE) module including a saliency erasing operation and a series of ordered learners. Specifically, for a specific frame of a video, the saliency erasing operation drives the specific learner to mine new and complementary parts by erasing the parts activated by previous frames. Such that the diverse visual features can be discovered for consecutive frames and finally form an integral characteristic of the target identity. Furthermore, a Temporal Saliency Boosting (TSB) module is designed to propagate the salient information among video frames to enhance the salient feature. It is complementary to TSE by effectively alleviating the information loss caused by the erasing operation of TSE. Extensive experiments show our method performs favorably against state-of-the-arts. The source code is available at https://github.com/blue-blue272/VideoReID-TCLNet.

Xinqian Gu, Hong Chang, Bingpeng Ma et al.

Due to the imperfect person detection results and posture changes, temporal appearance misalignment is unavoidable in video-based person re-identification (ReID). In this case, 3D convolution may destroy the appearance representation of person video clips, thus it is harmful to ReID. To address this problem, we propose AppearancePreserving 3D Convolution (AP3D), which is composed of two components: an Appearance-Preserving Module (APM) and a 3D convolution kernel. With APM aligning the adjacent feature maps in pixel level, the following 3D convolution can model temporal information on the premise of maintaining the appearance representation quality. It is easy to combine AP3D with existing 3D ConvNets by simply replacing the original 3D convolution kernels with AP3Ds. Extensive experiments demonstrate the effectiveness of AP3D for video-based ReID and the results on three widely used datasets surpass the state-of-the-arts. Code is available at: https://github.com/guxinqian/AP3D.

Yuhui Yuan, Jingyi Xie, Xilin Chen et al.

We present a model-agnostic post-processing scheme to improve the boundary quality for the segmentation result that is generated by any existing segmentation model. Motivated by the empirical observation that the label predictions of interior pixels are more reliable, we propose to replace the originally unreliable predictions of boundary pixels by the predictions of interior pixels. Our approach processes only the input image through two steps: (i) localize the boundary pixels and (ii) identify the corresponding interior pixel for each boundary pixel. We build the correspondence by learning a direction away from the boundary pixel to an interior pixel. Our method requires no prior information of the segmentation models and achieves nearly real-time speed. We empirically verify that our SegFix consistently reduces the boundary errors for segmentation results generated from various state-of-the-art models on Cityscapes, ADE20K and GTA5. Code is available at: https://github.com/openseg-group/openseg.pytorch.