arXiv subjects

Ruotong Wang

Publications and source records attributed to Ruotong Wang.

At least 19 recordsLinked to original sources

arXiv2026

Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM-Based Video Moderation

Despite their growing use in video moderation, multimodal large language models (MLLMs) exhibit a compositional safety blind spot: videos composed of seemingly benign components can convey harmful meaning when interpreted as a whole. We refer to this phenomenon as Distributed Implicit Harm (DIH), where harm arises from relations among components distributed along a decomposition axis of the video, rather than from any single explicit cue. Among many possible axes, we study two representative cases: temporally distributed harm across visual segments (DIH-T) and cross-modal harm between audio and visual streams (DIH-M). Studying and mitigating DIH at scale requires data that is difficult to collect: such videos lack compositional harm annotations, evade retrieval based on local visual cues, keywords, or single-modality signals, and are consequently absent from existing safety datasets. To bridge this gap, we develop a multi-agent synthesis framework that composes individually benign components into harmful scenarios and generates diverse DIH videos with explicit reasoning annotations, yielding a dataset of over 9,000 videos spanning visual-only and audio-visual settings. Benchmarking over 30 MLLMs spanning frontier proprietary models and leading open-source systems reveals substantial and consistent deficits in detecting both DIH-T and DIH-M. Notably, this failure persists even among the strongest frontier models: they often correctly assess individual components in isolation but fail to recognize the harmful meaning that emerges from their composition. We further evaluate these models on a manually collected set of real-world DIH videos from social media and observe the same failure mode, highlighting DIH as a practical and underexplored challenge for video moderation.

cs.CV↗

arXiv2026

BrandFusion: A Multi-Agent Framework for Seamless Brand Integration in Text-to-Video Generation

The rapid advancement of text-to-video (T2V) models has revolutionized content creation, yet their commercial potential remains largely untapped. We introduce, for the first time, the task of seamless brand integration in T2V: automatically embedding advertiser brands into prompt-generated videos while preserving semantic fidelity to user intent. This task confronts three core challenges: maintaining prompt fidelity, ensuring brand recognizability, and achieving contextually natural integration. To address them, we propose BrandFusion, a novel multi-agent framework comprising two synergistic phases. In the offline phase (advertiser-facing), we construct a Brand Knowledge Base by probing model priors and adapting to novel brands via lightweight fine-tuning. In the online phase (user-facing), five agents jointly refine user prompts through iterative refinement, leveraging the shared knowledge base and real-time contextual tracking to ensure brand visibility and semantic alignment. Experiments on 18 established and 2 custom brands across multiple state-of-the-art T2V models demonstrate that BrandFusion significantly outperforms baselines in semantic preservation, brand recognizability, and integration naturalness. Human evaluations further confirm higher user satisfaction, establishing a practical pathway for sustainable T2V monetization.

cs.CV↗

arXiv2025

Transferable Modeling Strategies for Low-Resource LLM Tasks: A Prompt and Alignment-Based Approach

This paper addresses the limited transfer and adaptation capabilities of large language models in low-resource language scenarios. It proposes a unified framework that combines a knowledge transfer module with parameter-efficient fine-tuning strategies. The method introduces knowledge alignment loss and soft prompt tuning to guide the model in effectively absorbing the structural features of target languages or tasks under minimal annotation. This enhances both generalization performance and training stability. The framework includes lightweight adaptation modules to reduce computational costs. During training, it integrates freezing strategies and prompt injection to preserve the model's original knowledge while enabling quick adaptation to new tasks. The study also conducts stability analysis experiments and synthetic pseudo-data transfer experiments to systematically evaluate the method's applicability and robustness across different low-resource tasks. Experimental results show that compared with existing multilingual pre-trained models and mainstream transfer methods, the proposed approach achieves higher performance and stability on cross-lingual tasks such as MLQA, XQuAD, and PAWS-X. It demonstrates particularly strong advantages under extremely data-scarce conditions. The proposed method offers strong generality and scalability. It enhances task-specific adaptability while preserving the general capabilities of large language models. This makes it well-suited for complex semantic modeling and multilingual processing tasks.

cs.CL↗

arXiv2025

BadVideo: Stealthy Backdoor Attack against Text-to-Video Generation

Text-to-video (T2V) generative models have rapidly advanced and found widespread applications across fields like entertainment, education, and marketing. However, the adversarial vulnerabilities of these models remain rarely explored. We observe that in T2V generation tasks, the generated videos often contain substantial redundant information not explicitly specified in the text prompts, such as environmental elements, secondary objects, and additional details, providing opportunities for malicious attackers to embed hidden harmful content. Exploiting this inherent redundancy, we introduce BadVideo, the first backdoor attack framework tailored for T2V generation. Our attack focuses on designing target adversarial outputs through two key strategies: (1) Spatio-Temporal Composition, which combines different spatiotemporal features to encode malicious information; (2) Dynamic Element Transformation, which introduces transformations in redundant elements over time to convey malicious information. Based on these strategies, the attacker's malicious target seamlessly integrates with the user's textual instructions, providing high stealthiness. Moreover, by exploiting the temporal dimension of videos, our attack successfully evades traditional content moderation systems that primarily analyze spatial information within individual frames. Extensive experiments demonstrate that BadVideo achieves high attack success rates while preserving original semantics and maintaining excellent performance on clean inputs. Overall, our work reveals the adversarial vulnerability of T2V models, calling attention to potential risks and misuse. Our project page is at https://wrt2000.github.io/BadVideo2025/.

cs.CV↗

arXiv2025

StructVPR++: Distill Structural and Semantic Knowledge with Weighting Samples for Visual Place Recognition

Visual place recognition is a challenging task for autonomous driving and robotics, which is usually considered as an image retrieval problem. A commonly used two-stage strategy involves global retrieval followed by re-ranking using patch-level descriptors. Most deep learning-based methods in an end-to-end manner cannot extract global features with sufficient semantic information from RGB images. In contrast, re-ranking can utilize more explicit structural and semantic information in one-to-one matching process, but it is time-consuming. To bridge the gap between global retrieval and re-ranking and achieve a good trade-off between accuracy and efficiency, we propose StructVPR++, a framework that embeds structural and semantic knowledge into RGB global representations via segmentation-guided distillation. Our key innovation lies in decoupling label-specific features from global descriptors, enabling explicit semantic alignment between image pairs without requiring segmentation during deployment. Furthermore, we introduce a sample-wise weighted distillation strategy that prioritizes reliable training pairs while suppressing noisy ones. Experiments on four benchmarks demonstrate that StructVPR++ surpasses state-of-the-art global methods by 5-23% in Recall@1 and even outperforms many two-stage approaches, achieving real-time efficiency with a single RGB input.

cs.CV↗

arXiv2025

To Think or Not to Think: Exploring the Unthinking Vulnerability in Large Reasoning Models

Large Reasoning Models (LRMs) are designed to solve complex tasks by generating explicit reasoning traces before producing final answers. However, we reveal a critical vulnerability in LRMs -- termed Unthinking Vulnerability -- wherein the thinking process can be bypassed by manipulating special delimiter tokens. It is empirically demonstrated to be widespread across mainstream LRMs, posing both a significant risk and potential utility, depending on how it is exploited. In this paper, we systematically investigate this vulnerability from both malicious and beneficial perspectives. On the malicious side, we introduce Breaking of Thought (BoT), a novel attack that enables adversaries to bypass the thinking process of LRMs, thereby compromising their reliability and availability. We present two variants of BoT: a training-based version that injects backdoor during the fine-tuning stage, and a training-free version based on adversarial attack during the inference stage. As a potential defense, we propose thinking recovery alignment to partially mitigate the vulnerability. On the beneficial side, we introduce Monitoring of Thought (MoT), a plug-and-play framework that allows model owners to enhance efficiency and safety. It is implemented by leveraging the same vulnerability to dynamically terminate redundant or risky reasoning through external monitoring. Extensive experiments show that BoT poses a significant threat to reasoning reliability, while MoT provides a practical solution for preventing overthinking and jailbreaking. Our findings expose an inherent flaw in current LRM architectures and underscore the need for more robust reasoning systems in the future.

cs.CL↗

arXiv2024

Social-RAG: Retrieving from Group Interactions to Socially Ground AI Generation

AI agents are increasingly tasked with making proactive suggestions in online spaces where groups collaborate, yet risk being unhelpful or even annoying if they fail to match group preferences or behave in socially inappropriate ways. Fortunately, group spaces have a rich history of prior interactions and affordances for social feedback that can support grounding an agent's generations to a group's interests and norms. We present Social-RAG, a workflow for socially grounding agents that retrieves context from prior group interactions, selects relevant social signals, and feeds them into a language model to generate messages in a socially aligned manner. We implement this in \textsc{PaperPing}, a system for posting paper recommendations in group chat, leveraging social signals determined from formative studies with 39 researchers. From a three-month deployment in 18 channels reaching 500+ researchers, we observed PaperPing posted relevant messages in groups without disrupting their existing social practices, fostering group common ground.

cs.HC↗

arXiv2024

Automated Genre-Aware Article Scoring and Feedback Using Large Language Models

This paper focuses on the development of an advanced intelligent article scoring system that not only assesses the overall quality of written work but also offers detailed feature-based scoring tailored to various article genres. By integrating the pre-trained BERT model with the large language model Chat-GPT, the system gains a deep understanding of both the content and structure of the text, enabling it to provide a thorough evaluation along with targeted suggestions for improvement. Experimental results demonstrate that this system outperforms traditional scoring methods across multiple public datasets, particularly in feature-based assessments, offering a more accurate reflection of the quality of different article types. Moreover, the system generates personalized feedback to assist users in enhancing their writing skills, underscoring the potential and practical value of automated scoring technologies in educational contexts.

cs.CL↗

arXiv2024

Integration of Mamba and Transformer -- MAT for Long-Short Range Time Series Forecasting with Application to Weather Dynamics

Long-short range time series forecasting is essential for predicting future trends and patterns over extended periods. While deep learning models such as Transformers have made significant strides in advancing time series forecasting, they often encounter difficulties in capturing long-term dependencies and effectively managing sparse semantic features. The state-space model, Mamba, addresses these issues through its adept handling of selective input and parallel computing, striking a balance between computational efficiency and prediction accuracy. This article examines the advantages and disadvantages of both Mamba and Transformer models, and introduces a combined approach, MAT, which leverages the strengths of each model to capture unique long-short range dependencies and inherent evolutionary patterns in multivariate time series. Specifically, MAT harnesses the long-range dependency capabilities of Mamba and the short-range characteristics of Transformers. Experimental results on benchmark weather datasets demonstrate that MAT outperforms existing comparable methods in terms of prediction accuracy, scalability, and memory efficiency.

cs.LG↗

arXiv2024

BackdoorBench: A Comprehensive Benchmark and Analysis of Backdoor Learning

As an emerging and vital topic for studying deep neural networks' vulnerability (DNNs), backdoor learning has attracted increasing interest in recent years, and many seminal backdoor attack and defense algorithms are being developed successively or concurrently, in the status of a rapid arms race. However, mainly due to the diverse settings, and the difficulties of implementation and reproducibility of existing works, there is a lack of a unified and standardized benchmark of backdoor learning, causing unfair comparisons, and unreliable conclusions (e.g., misleading, biased or even false conclusions). Consequently, it is difficult to evaluate the current progress and design the future development roadmap of this literature. To alleviate this dilemma, we build a comprehensive benchmark of backdoor learning called BackdoorBench. Our benchmark makes three valuable contributions to the research community. 1) We provide an integrated implementation of state-of-the-art (SOTA) backdoor learning algorithms (currently including 16 attack and 27 defense algorithms), based on an extensible modular-based codebase. 2) We conduct comprehensive evaluations of 12 attacks against 16 defenses, with 5 poisoning ratios, based on 4 models and 4 datasets, thus 11,492 pairs of evaluations in total. 3) Based on above evaluations, we present abundant analysis from 8 perspectives via 18 useful analysis tools, and provide several inspiring insights about backdoor learning. We hope that our efforts could build a solid foundation of backdoor learning to facilitate researchers to investigate existing algorithms, develop more innovative algorithms, and explore the intrinsic mechanism of backdoor learning. Finally, we have created a user-friendly website at http://backdoorbench.com, which collects all important information of BackdoorBench, including codebase, docs, leaderboard, and model Zoo.

cs.CV↗

arXiv2024

BackdoorBench: A Comprehensive Benchmark and Analysis of Backdoor Learning

As an emerging approach to explore the vulnerability of deep neural networks (DNNs), backdoor learning has attracted increasing interest in recent years, and many seminal backdoor attack and defense algorithms are being developed successively or concurrently, in the status of a rapid arms race. However, mainly due to the diverse settings, and the difficulties of implementation and reproducibility of existing works, there is a lack of a unified and standardized benchmark of backdoor learning, causing unfair comparisons or unreliable conclusions (e.g., misleading, biased or even false conclusions). Consequently, it is difficult to evaluate the current progress and design the future development roadmap of this literature. To alleviate this dilemma, we build a comprehensive benchmark of backdoor learning called BackdoorBench. Our benchmark makes three valuable contributions to the research community. 1) We provide an integrated implementation of state-of-the-art (SOTA) backdoor learning algorithms (currently including 20 attack and 32 defense algorithms), based on an extensible modular-based codebase. 2) We conduct comprehensive evaluations with 5 poisoning ratios, based on 4 models and 4 datasets, leading to 11,492 pairs of attack-against-defense evaluations in total. 3) Based on above evaluations, we present abundant analysis from 10 perspectives via 18 useful analysis tools, and provide several inspiring insights about backdoor learning. We hope that our efforts could build a solid foundation of backdoor learning to facilitate researchers to investigate existing algorithms, develop more innovative algorithms, and explore the intrinsic mechanism of backdoor learning. Finally, we have created a user-friendly website at http://backdoorbench.com, which collects all important information of BackdoorBench, including codebase, docs, leaderboard, and model Zoo.

cs.LG↗

arXiv2024

Versatile Backdoor Attack with Visible, Semantic, Sample-Specific, and Compatible Triggers

Deep neural networks (DNNs) can be manipulated to exhibit specific behaviors when exposed to specific trigger patterns, without affecting their performance on benign samples, dubbed \textit{backdoor attack}. Currently, implementing backdoor attacks in physical scenarios still faces significant challenges. Physical attacks are labor-intensive and time-consuming, and the triggers are selected in a manual and heuristic way. Moreover, expanding digital attacks to physical scenarios faces many challenges due to their sensitivity to visual distortions and the absence of counterparts in the real world. To address these challenges, we define a novel trigger called the \textbf{V}isible, \textbf{S}emantic, \textbf{S}ample-Specific, and \textbf{C}ompatible (VSSC) trigger, to achieve effective, stealthy and robust simultaneously, which can also be effectively deployed in the physical scenario using corresponding objects. To implement the VSSC trigger, we propose an automated pipeline comprising three modules: a trigger selection module that systematically identifies suitable triggers leveraging large language models, a trigger insertion module that employs generative models to seamlessly integrate triggers into images, and a quality assessment module that ensures the natural and successful insertion of triggers through vision-language models. Extensive experimental results and analysis validate the effectiveness, stealthiness, and robustness of the VSSC trigger. It can not only maintain robustness under visual distortions but also demonstrates strong practicality in the physical scenario. We hope that the proposed VSSC trigger and implementation approach could inspire future studies on designing more practical triggers in backdoor attacks.

cs.CV↗

arXiv2024

Investigating and Designing for Trust in AI-powered Code Generation Tools

As AI-powered code generation tools such as GitHub Copilot become popular, it is crucial to understand software developers' trust in AI tools -- a key factor for tool adoption and responsible usage. However, we know little about how developers build trust with AI, nor do we understand how to design the interface of generative AI systems to facilitate their appropriate levels of trust. In this paper, we describe findings from a two-stage qualitative investigation. We first interviewed 17 developers to contextualize their notions of trust and understand their challenges in building appropriate trust in AI code generation tools. We surfaced three main challenges -- including building appropriate expectations, configuring AI tools, and validating AI suggestions. To address these challenges, we conducted a design probe study in the second stage to explore design concepts that support developers' trust-building process by 1) communicating AI performance to help users set proper expectations, 2) allowing users to configure AI by setting and adjusting preferences, and 3) offering indicators of model mechanism to support evaluation of AI suggestions. We gathered developers' feedback on how these design concepts can help them build appropriate trust in AI-powered code generation tools, as well as potential risks in design. These findings inform our proposed design recommendations on how to design for trust in AI-powered code generation tools.

cs.HC↗

arXiv2024

Meeting Bridges: Designing Information Artifacts that Bridge from Synchronous Meetings to Asynchronous Collaboration

A recent surge in remote meetings has led to complaints of ``Zoom fatigue'' and ``collaboration overload,'' negatively impacting worker productivity and well-being. One way to alleviate the burden of meetings is to de-emphasize their synchronous participation by shifting work to and enabling sensemaking during post-meeting asynchronous activities. Towards this goal, we propose the design concept of meeting bridges, or information artifacts that can encapsulate meeting information towards bridging to and facilitating post-meeting activities. Through 13 interviews and a survey of 198 information workers, we learn how people use online meeting information after meetings are over, finding five main uses: as an archive, as task reminders, to onboard or support inclusion, for group sensemaking, and as a launching point for follow-on collaboration. However, we also find that current common meeting artifacts, such as notes and recordings, present challenges in serving as meeting bridges. After conducting co-design sessions with 16 participants, we distill key principles for the design of meeting bridges to optimally support asynchronous collaboration goals. Overall, our findings point to the opportunity of designing information artifacts that not only support users to access but also continue to transform and engage in meeting information post-meeting.

cs.HC↗

arXiv2023

"Is Reporting Worth the Sacrifice of Revealing What I Have Sent?": Privacy Considerations When Reporting on End-to-End Encrypted Platforms

User reporting is an essential component of content moderation on many online platforms -- in particular, on end-to-end encrypted (E2EE) messaging platforms where platform operators cannot proactively inspect message contents. However, users' privacy concerns when considering reporting may impede the effectiveness of this strategy in regulating online harassment. In this paper, we conduct interviews with 16 users of E2EE platforms to understand users' mental models of how reporting works and their resultant privacy concerns and considerations surrounding reporting. We find that users expect platforms to store rich longitudinal reporting datasets, recognizing both their promise for better abuse mitigation and the privacy risk that platforms may exploit or fail to protect them. We also find that users have preconceptions about the respective capabilities and risks of moderators at the platform versus community level -- for instance, users trust platform moderators more to not abuse their power but think community moderators have more time to attend to reports. These considerations, along with perceived effectiveness of reporting and how to provide sufficient evidence while maintaining privacy, shape how users decide whether, to whom, and how much to report. We conclude with design implications for a more privacy-preserving reporting system on E2EE messaging platforms.

cs.HC↗

arXiv2023

Personalizing Content Moderation on Social Media: User Perspectives on Moderation Choices, Interface Design, and Labor

Social media platforms moderate content for each user by incorporating the outputs of both platform-wide content moderation systems and, in some cases, user-configured personal moderation preferences. However, it is unclear (1) how end users perceive the choices and affordances of different kinds of personal content moderation tools, and (2) how the introduction of personalization impacts user perceptions of platforms' content moderation responsibilities. This paper investigates end users' perspectives on personal content moderation tools by conducting an interview study with a diverse sample of 24 active social media users. We probe interviewees' preferences using simulated personal moderation interfaces, including word filters, sliders for toxicity levels, and boolean toxicity toggles. We also examine the labor involved for users in choosing moderation settings and present users' attitudes about the roles and responsibilities of social media platforms and other stakeholders towards moderation. We discuss how our findings can inform design solutions to improve transparency and controllability in personal content moderation tools.

cs.HC↗

arXiv2023

"It would work for me too": How Online Communities Shape Software Developers' Trust in AI-Powered Code Generation Tools

While revolutionary AI-powered code generation tools have been rising rapidly, we know little about how and how to help software developers form appropriate trust in those AI tools. Through a two-phase formative study, we investigate how online communities shape developers' trust in AI tools and how we can leverage community features to facilitate appropriate user trust. Through interviewing 17 developers, we find that developers collectively make sense of AI tools using the experiences shared by community members and leverage community signals to evaluate AI suggestions. We then surface design opportunities and conduct 11 design probe sessions to explore the design space of using community features to support user trust in AI code generation systems. We synthesize our findings and extend an existing model of user trust in AI technologies with sociotechnical factors. We map out the design considerations for integrating user community into the AI code generation experience.

cs.HC↗

arXiv2023

StructVPR: Distill Structural Knowledge with Weighting Samples for Visual Place Recognition

Visual place recognition (VPR) is usually considered as a specific image retrieval problem. Limited by existing training frameworks, most deep learning-based works cannot extract sufficiently stable global features from RGB images and rely on a time-consuming re-ranking step to exploit spatial structural information for better performance. In this paper, we propose StructVPR, a novel training architecture for VPR, to enhance structural knowledge in RGB global features and thus improve feature stability in a constantly changing environment. Specifically, StructVPR uses segmentation images as a more definitive source of structural knowledge input into a CNN network and applies knowledge distillation to avoid online segmentation and inference of seg-branch in testing. Considering that not all samples contain high-quality and helpful knowledge, and some even hurt the performance of distillation, we partition samples and weigh each sample's distillation loss to enhance the expected knowledge precisely. Finally, StructVPR achieves impressive performance on several benchmarks using only global retrieval and even outperforms many two-stage approaches by a large margin. After adding additional re-ranking, ours achieves state-of-the-art performance while maintaining a low computational cost.

cs.CV↗