arXiv subjects

Shiqing Ma

Publications and source records attributed to Shiqing Ma.

At least 37 records · Page 2Linked to original sources

arXiv2025

VIDSTAMP: A Temporally-Aware Watermark for Ownership and Integrity in Video Diffusion Models

Video diffusion models can generate realistic and temporally consistent videos. This raises concerns about provenance, ownership, and integrity. Watermarking can help address these issues by embedding metadata directly into the content. To work well, a watermark needs enough capacity for meaningful metadata. It must also stay imperceptible and remain robust to common video manipulations. Existing methods struggle with limited capacity, extra inference cost, or reduced visual quality. We introduce VidStamp, a watermarking framework that embeds frame-level messages through the decoder of a latent video diffusion model. The decoder is fine-tuned in two stages. The first stage uses static image datasets to encourage spatial message separation. The second stage uses synthesized video sequences to restore temporal consistency. This approach enables high-capacity watermarks with minimal perceptual impact. VidStamp also supports dynamic watermarking through a control signal that selects message templates during inference. This adds flexibility and creates a second channel for communication. We evaluate VidStamp on Stable Video Diffusion (I2V), OpenSora, and Wan (T2V). The system embeds 48 bits per frame while preserving visual quality and staying robust to common distortions. Compared with VideoSeal, VideoShield, and RivaGAN, it achieves lower log P-values and stronger detectability. Its frame-wise watermarking design also enables precise temporal tamper localization, with an accuracy of 0.96, which exceeds the VideoShield baseline. Code: https://github.com/SPIN-UMass/VidStamp

cs.CV↗

arXiv2025

CITADEL: Context Similarity Based Deep Learning Framework Bug Finding

With the application of deep learning technology, tools of DL framework testing are in high demand. Existing DL framework testing tools have limited coverage of bug types. For example, they lack the capability of effectively finding performance bugs, which are critical for DL models regarding performance, economics, and the environment. Moreover, existing tools are inefficient, generating hundreds of test cases with few trigger bugs. In this paper, we propose Citadel, a method that accelerates bug finding in terms of efficiency and effectiveness. We observe that many DL framework bugs are similar due to the similarity of operators and algorithms belonging to the same family. Orthogonal to existing bug-finding tools, Citadel aims to find new bugs that are similar to reported ones that have known test oracles. Citadel defines context similarity to measure the similarity of DL framework API pairs and automatically generates test cases with oracles for APIs that are similar to the problematic APIs in existing bug reports. Citadel effectively detects 58 and 66 API bugs on PyTorch and TensorFlow (excluding those rejected by developers or duplicates of prior reports), many of which, e.g., 13 performance bugs, cannot be detected by existing tools. Moreover, 35.40% of test cases generated by Citadel can trigger bugs significantly transcending the state-of-the-art method (3.90%).

cs.SE↗

arXiv2025

MCP-RADAR: A Multi-Dimensional Benchmark for Evaluating Tool Use Capabilities in Large Language Models

As Large Language Models (LLMs) evolve from passive text generators to active reasoning agents capable of interacting with external tools, the Model Context Protocol (MCP) has emerged as a key standardized framework for dynamic tool discovery and orchestration. Despite its widespread industry adoption, existing evaluation methods do not adequately assess tool utilization capabilities under this new paradigm. To address this gap, this paper introduces MCP-RADAR, the first comprehensive benchmark specifically designed to evaluate LLM performance within the MCP framework. MCP-RADAR features a challenging dataset of 507 tasks spanning six domains: mathematical reasoning, web search, email, calendar, file management, and terminal operations. It quantifies performance based on two primary criteria: answer correctness and operational accuracy. To closely emulate real-world usage, our evaluation employs both authentic MCP tools and high-fidelity simulations of official tools. Unlike traditional benchmarks that rely on subjective human evaluation or binary success metrics, MCP-RADAR adopts objective, quantifiable measurements across multiple task domains, including computational resource efficiency and the number of successful tool-invocation rounds. Our evaluation of leading closed-source and open-source LLMs reveals distinct capability profiles and highlights a significant trade-off between accuracy and efficiency. Our findings provide actionable insights for both LLM developers and tool creators, establishing a standardized methodology applicable to the broader LLM agent ecosystem. All implementations, configurations, and datasets are publicly available at https://anonymous.4open.science/r/MCPRadar-B143.

cs.AI↗

arXiv2025

Rethinking Provenance Completeness with a Learning-Based Linux Scheduler

Provenance plays a critical role in maintaining traceability of a system's actions for root cause analysis of security threats and impacts. Provenance collection is often incorporated into the reference monitor of systems to ensure that an audit trail exists of all events, that events are completely captured, and that logging of such events cannot be bypassed. However, recent research has questioned whether existing state-of-the-art provenance collection systems fail to ensure the security guarantees of a true reference monitor due to the 'super producer threat' in which provenance generation can overload a system to force the system to drop security-relevant events and allow an attacker to hide their actions. One approach towards solving this threat is to enforce resource isolation, but that does not fully solve the problems resulting from hardware dependencies and performance limitations. In this paper, we show how an operating system's kernel scheduler can mitigate this threat, and we introduce Aegis, a learned scheduler for Linux specifically designed for provenance. Unlike conventional schedulers that ignore provenance completeness requirements, Aegis leverages reinforcement learning to learn provenance task behavior and to dynamically optimize resource allocation. We evaluate Aegis's efficacy and show that Aegis significantly improves both the completeness and efficiency of provenance collection systems compared to traditional scheduling, while maintaining reasonable overheads and even improving overall runtime in certain cases compared to the default Linux scheduler.

cs.CR↗

arXiv2025

MeanSparse: Post-Training Robustness Enhancement Through Mean-Centered Feature Sparsification

We present a simple yet effective method to improve the robustness of both Convolutional and attention-based Neural Networks against adversarial examples by post-processing an adversarially trained model. Our technique, MeanSparse, cascades the activation functions of a trained model with novel operators that sparsify mean-centered feature vectors. This is equivalent to reducing feature variations around the mean, and we show that such reduced variations merely affect the model's utility, yet they strongly attenuate the adversarial perturbations and decrease the attacker's success rate. Our experiments show that, when applied to the top models in the RobustBench leaderboard, MeanSparse achieves a new robustness record of 75.28% (from 73.71%), 44.78% (from 42.67%) and 62.12% (from 59.56%) on CIFAR-10, CIFAR-100 and ImageNet, respectively, in terms of AutoAttack accuracy. Code is available at https://github.com/SPIN-UMass/MeanSparse

cs.CV↗

arXiv2025

Bias Similarity Measurement: A Black-Box Audit of Fairness Across LLMs

Large Language Models (LLMs) reproduce social biases, yet prevailing evaluations score models in isolation, obscuring how biases persist across families and releases. We introduce Bias Similarity Measurement (BSM), which treats fairness as a relational property between models, unifying scalar, distributional, behavioral, and representational signals into a single similarity space. Evaluating 30 LLMs on 1M+ prompts, we find that instruction tuning primarily enforces abstention rather than altering internal representations; small models gain little accuracy and can become less fair under forced choice; and open-weight models can match or exceed proprietary systems. Family signatures diverge: Gemma favors refusal, LLaMA 3.1 approaches neutrality with fewer refusals, and converges toward abstention-heavy behavior overall. Counterintuitively, Gemma 3 Instruct matches GPT-4-level fairness at far lower cost, whereas Gemini's heavy abstention suppresses utility. Beyond these findings, BSM offers an auditing workflow for procurement, regression testing, and lineage screening, and extends naturally to code and multilingual settings. Our results reframe fairness not as isolated scores but as comparative bias similarity, enabling systematic auditing of LLM ecosystems. Code available at https://github.com/HyejunJeong/bias_llm.

cs.LG↗

arXiv2025

DeCoMa: Detecting and Purifying Code Dataset Watermarks through Dual Channel Code Abstraction

Watermarking is a technique to help identify the source of data points, which can be used to help prevent the misuse of protected datasets. Existing methods on code watermarking, leveraging the idea from the backdoor research, embed stealthy triggers as watermarks. Despite their high resilience against dilution attacks and backdoor detections, the robustness has not been fully evaluated. To fill this gap, we propose DeCoMa, a dual-channel approach to Detect and purify Code dataset waterMarks. To overcome the high barrier created by the stealthy and hidden nature of code watermarks, DeCoMa leverages dual-channel constraints on code to generalize and map code samples into standardized templates. Subsequently, DeCoMa extracts hidden watermarks by identifying outlier associations between paired elements within the standardized templates. Finally, DeCoMa purifies the watermarked dataset by removing all samples containing the detected watermark, enabling the silent appropriation of protected code. We conduct extensive experiments to evaluate the effectiveness and efficiency of DeCoMa, covering 14 types of code watermarks and 3 representative intelligent code tasks (a total of 14 scenarios). Experimental results demonstrate that DeCoMa achieves a stable recall of 100% in 14 code watermark detection scenarios, significantly outperforming the baselines. Additionally, DeCoMa effectively attacks code watermarks with embedding rates as low as 0.1%, while maintaining comparable model performance after training on the purified dataset. Furthermore, as DeCoMa requires no model training for detection, it achieves substantially higher efficiency than all baselines, with a speedup ranging from 31.5 to 130.9X. The results call for more advanced watermarking techniques for code models, while DeCoMa can serve as a baseline for future evaluation. Code is available at https://github.com/xiaoyuanpigo/DeCoMa

cs.CR↗

arXiv2025

Mitigating Stylistic Biases of Machine Translation Systems via Monolingual Corpora Only

The advent of neural machine translation (NMT) has revolutionized cross-lingual communication, yet preserving stylistic nuances remains a significant challenge. While existing approaches often require parallel corpora for style preservation, we introduce Babel, a novel framework that enhances stylistic fidelity in NMT using only monolingual corpora. Babel employs two key components: (1) a style detector based on contextual embeddings that identifies stylistic disparities between source and target texts, and (2) a diffusion-based style applicator that rectifies stylistic inconsistencies while maintaining semantic integrity. Our framework integrates with existing NMT systems as a post-processing module, enabling style-aware translation without requiring architectural modifications or parallel stylistic data. Extensive experiments on five diverse domains (law, literature, scientific writing, medicine, and educational content) demonstrate Babel's effectiveness: it identifies stylistic inconsistencies with 88.21% precision and improves stylistic preservation by 150% while maintaining a high semantic similarity score of 0.92. Human evaluation confirms that translations refined by Babel better preserve source text style while maintaining fluency and adequacy.

cs.CL↗

arXiv2025

ASSURE: Metamorphic Testing for AI-powered Browser Extensions

The integration of Large Language Models (LLMs) into browser extensions has revolutionized web browsing, enabling sophisticated functionalities like content summarization, intelligent translation, and context-aware writing assistance. However, these AI-powered extensions introduce unprecedented challenges in testing and reliability assurance. Traditional browser extension testing approaches fail to address the non-deterministic behavior, context-sensitivity, and complex web environment integration inherent to LLM-powered extensions. Similarly, existing LLM testing methodologies operate in isolation from browser-specific contexts, creating a critical gap in effective evaluation frameworks. To bridge this gap, we present ASSURE, a modular automated testing framework specifically designed for AI-powered browser extensions. ASSURE comprises three principal components: (1) a modular test case generation engine that supports plugin-based extension of testing scenarios, (2) an automated execution framework that orchestrates the complex interactions between web content, extension processing, and AI model behavior, and (3) a configurable validation pipeline that systematically evaluates behavioral consistency and security invariants rather than relying on exact output matching. Our evaluation across six widely-used AI browser extensions demonstrates ASSURE's effectiveness, identifying 531 distinct issues spanning security vulnerabilities, metamorphic relation violations, and content alignment problems. ASSURE achieves 6.4x improved testing throughput compared to manual approaches, detecting critical security vulnerabilities within 12.4 minutes on average. This efficiency makes ASSURE practical for integration into development pipelines, offering a comprehensive solution to the unique challenges of testing AI-powered browser extensions.

cs.SE↗

arXiv2025

The Foundation Cracks: A Comprehensive Study on Bugs and Testing Practices in LLM Libraries

Large Language Model (LLM) libraries have emerged as the foundational infrastructure powering today's AI revolution, serving as the backbone for LLM deployment, inference optimization, fine-tuning, and production serving across diverse applications. Despite their critical role in the LLM ecosystem, these libraries face frequent quality issues and bugs that threaten the reliability of AI systems built upon them. To address this knowledge gap, we present the first comprehensive empirical investigation into bug characteristics and testing practices in modern LLM libraries. We examine 313 bug-fixing commits extracted across two widely-adopted LLM libraries: HuggingFace Transformers and vLLM.Through rigorous manual analysis, we establish comprehensive taxonomies categorizing bug symptoms into 5 types and root causes into 14 distinct categories.Our primary discovery shows that API misuse has emerged as the predominant root cause (32.17%-48.19%), representing a notable transition from algorithm-focused defects in conventional deep learning frameworks toward interface-oriented problems. Additionally, we examine 7,748 test functions to identify 7 distinct test oracle categories employed in current testing approaches, with predefined expected outputs (such as specific tensors and text strings) being the most common strategy. Our assessment of existing testing effectiveness demonstrates that the majority of bugs escape detection due to inadequate test cases (41.73%), lack of test drivers (32.37%), and weak test oracles (25.90%). Drawing from these findings, we offer some recommendations for enhancing LLM library quality assurance.

cs.SE↗

arXiv2025

The Invisible Hand: Unveiling Provider Bias in Large Language Models for Code Generation

Large Language Models (LLMs) have emerged as the new recommendation engines, surpassing traditional methods in both capability and scope, particularly in code generation. In this paper, we reveal a novel provider bias in LLMs: without explicit directives, these models show systematic preferences for services from specific providers in their recommendations (e.g., favoring Google Cloud over Microsoft Azure). To systematically investigate this bias, we develop an automated pipeline to construct the dataset, incorporating 6 distinct coding task categories and 30 real-world application scenarios. Leveraging this dataset, we conduct the first comprehensive empirical study of provider bias in LLM code generation across seven state-of-the-art LLMs, utilizing approximately 500 million tokens (equivalent to $5,000+ in computational costs). Our findings reveal that LLMs exhibit significant provider preferences, predominantly favoring services from Google and Amazon, and can autonomously modify input code to incorporate their preferred providers without users' requests. Such a bias holds far-reaching implications for market dynamics and societal equilibrium, potentially contributing to digital monopolies. It may also deceive users and violate their expectations, leading to various consequences. We call on the academic community to recognize this emerging issue and develop effective evaluation and mitigation methods to uphold AI security and fairness.

cs.SE↗

arXiv2025

Token-Budget-Aware LLM Reasoning

Reasoning is critical for large language models (LLMs) to excel in a wide range of tasks. While methods like Chain-of-Thought (CoT) reasoning and enhance LLM performance by decomposing problems into intermediate steps, they also incur significant overhead in token usage, leading to increased costs. We find that the reasoning process of current LLMs is unnecessarily lengthy and it can be compressed by including a reasonable token budget in the prompt, but the choice of token budget plays a crucial role in the actual compression effectiveness. We then propose a token-budget-aware LLM reasoning framework that dynamically adjusts the number of reasoning tokens based on the reasoning complexity of each problem. Experiments show that our method effectively reduces token costs in CoT reasoning with only a slight performance reduction, offering a practical solution to balance efficiency and accuracy in LLM reasoning. Code: https://github.com/GeniusHTX/TALE

cs.CL↗

arXiv2025

From Effectiveness to Efficiency: Uncovering Linguistic Bias in Large Language Model-based Code Generation

Large Language Models (LLMs) have demonstrated promising capabilities for code generation. While existing benchmarks evaluate the correctness and efficiency of LLM-generated code, the potential linguistic bias - where code quality varies based on the natural language used to describe programming tasks - remains underexplored. In this paper, we aim to investigate this linguistic bias through the lens of English and Chinese. To facilitate our investigation, we present a unified evaluation framework comprising a curated dataset of 52 Python programming questions with parallel bilingual task descriptions, automated correctness verification, and efficiency quantification tools based on runtime complexity estimation. Based on this framework, we conduct the first empirical study towards the linguistic bias in LLM-generated code on eight popular LCGMs, as well as GPT-3.5-Turbo and GPT-4. We observe that these LCGM-generated code show different correctness on an average of 12% bilingual programming tasks, where 39% also exhibits diverse efficiency. Our findings indicate that LLMs commonly exhibit linguistic bias for code generation.

cs.SE↗

arXiv2025

A Survey on Federated Unlearning: Challenges and Opportunities

Federated learning (FL), introduced in 2017, facilitates collaborative learning between non-trusting parties with no need for the parties to explicitly share their data among themselves. This allows training models on user data while respecting privacy regulations such as GDPR and CPRA. However, emerging privacy requirements may mandate model owners to be able to \emph{forget} some learned data, e.g., when requested by data owners or law enforcement. This has given birth to an active field of research called \emph{machine unlearning}. In the context of FL, many techniques developed for unlearning in centralized settings are not trivially applicable! This is due to the unique differences between centralized and distributed learning, in particular, interactivity, stochasticity, heterogeneity, and limited accessibility in FL. In response, a recent line of work has focused on developing unlearning mechanisms tailored to FL. This SoK paper aims to take a deep look at the \emph{federated unlearning} literature, with the goal of identifying research trends and challenges in this emerging field. By carefully categorizing papers published on FL unlearning (since 2020), we aim to pinpoint the unique complexities of federated unlearning, highlighting limitations on directly applying centralized unlearning methods. We compare existing federated unlearning methods regarding influence removal and performance recovery, compare their threat models and assumptions, and discuss their implications and limitations. For instance, we analyze the experimental setup of FL unlearning studies from various perspectives, including data heterogeneity and its simulation, the datasets used for demonstration, and evaluation metrics. Our work aims to offer insights and suggestions for future research on federated unlearning.

cs.LG↗

arXiv2025

MLLM-as-a-Judge for Image Safety without Human Labeling

Image content safety has become a significant challenge with the rise of visual media on online platforms. Meanwhile, in the age of AI-generated content (AIGC), many image generation models are capable of producing harmful content, such as images containing sexual or violent material. Thus, it becomes crucial to identify such unsafe images based on established safety rules. Pre-trained Multimodal Large Language Models (MLLMs) offer potential in this regard, given their strong pattern recognition abilities. Existing approaches typically fine-tune MLLMs with human-labeled datasets, which however brings a series of drawbacks. First, relying on human annotators to label data following intricate and detailed guidelines is both expensive and labor-intensive. Furthermore, users of safety judgment systems may need to frequently update safety rules, making fine-tuning on human-based annotation more challenging. This raises the research question: Can we detect unsafe images by querying MLLMs in a zero-shot setting using a predefined safety constitution (a set of safety rules)? Our research showed that simply querying pre-trained MLLMs does not yield satisfactory results. This lack of effectiveness stems from factors such as the subjectivity of safety rules, the complexity of lengthy constitutions, and the inherent biases in the models. To address these challenges, we propose a MLLM-based method includes objectifying safety rules, assessing the relevance between rules and images, making quick judgments based on debiased token probabilities with logically complete yet simplified precondition chains for safety rules, and conducting more in-depth reasoning with cascaded chain-of-thought processes if necessary. Experiment results demonstrate that our method is highly effective for zero-shot image safety judgment tasks.

cs.CV↗

arXiv2025

Invisible Backdoor Attack against Self-supervised Learning

Self-supervised learning (SSL) models are vulnerable to backdoor attacks. Existing backdoor attacks that are effective in SSL often involve noticeable triggers, like colored patches or visible noise, which are vulnerable to human inspection. This paper proposes an imperceptible and effective backdoor attack against self-supervised models. We first find that existing imperceptible triggers designed for supervised learning are less effective in compromising self-supervised models. We then identify this ineffectiveness is attributed to the overlap in distributions between the backdoor and augmented samples used in SSL. Building on this insight, we design an attack using optimized triggers disentangled with the augmented transformation in the SSL, while remaining imperceptible to human vision. Experiments on five datasets and six SSL algorithms demonstrate our attack is highly effective and stealthy. It also has strong resistance to existing backdoor defenses. Our code can be found at https://github.com/Zhang-Henry/INACTIVE.

cs.CV↗

arXiv2025

An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer

Despite prior safety alignment efforts, mainstream LLMs can still generate harmful and unethical content when subjected to jailbreaking attacks. Existing jailbreaking methods fall into two main categories: template-based and optimization-based methods. The former requires significant manual effort and domain knowledge, while the latter, exemplified by Greedy Coordinate Gradient (GCG), which seeks to maximize the likelihood of harmful LLM outputs through token-level optimization, also encounters several limitations: requiring white-box access, necessitating pre-constructed affirmative phrase, and suffering from low efficiency. In this paper, we present ECLIPSE, a novel and efficient black-box jailbreaking method utilizing optimizable suffixes. Drawing inspiration from LLMs' powerful generation and optimization capabilities, we employ task prompts to translate jailbreaking goals into natural language instructions. This guides the LLM to generate adversarial suffixes for malicious queries. In particular, a harmfulness scorer provides continuous feedback, enabling LLM self-reflection and iterative optimization to autonomously and efficiently produce effective suffixes. Experimental results demonstrate that ECLIPSE achieves an average attack success rate (ASR) of 0.92 across three open-source LLMs and GPT-3.5-Turbo, significantly surpassing GCG in 2.4 times. Moreover, ECLIPSE is on par with template-based methods in ASR while offering superior attack efficiency, reducing the average attack overhead by 83%.

cs.AI↗

arXiv2025

Tightening Robustness Verification of MaxPool-based Neural Networks via Minimizing the Over-Approximation Zone

The robustness of neural network classifiers is important in the safety-critical domain and can be quantified by robustness verification. At present, efficient and scalable verification techniques are always sound but incomplete, and thus, the improvement of verified robustness results is the key criterion to evaluate the performance of incomplete verification approaches. The multi-variate function MaxPool is widely adopted yet challenging to verify. In this paper, we present Ti-Lin, a robustness verifier for MaxPool-based CNNs with Tight Linear Approximation. Following the sequel of minimizing the over-approximation zone of the non-linear function of CNNs, we are the first to propose the provably neuron-wise tightest linear bounds for the MaxPool function. By our proposed linear bounds, we can certify larger robustness results for CNNs. We evaluate the effectiveness of Ti-Lin on different verification frameworks with open-sourced benchmarks, including LeNet, PointNet, and networks trained on the MNIST, CIFAR-10, Tiny ImageNet and ModelNet40 datasets. Experimental results show that Ti-Lin significantly outperforms the state-of-the-art methods across all networks with up to 78.6% improvement in terms of the certified accuracy with almost the same time consumption as the fastest tool. Our code is available at https://github.com/xiaoyuanpigo/Ti-Lin-Hybrid-Lin.

cs.LG↗