arXiv subjects

Jingling Xue

Publications and source records attributed to Jingling Xue.

At least 19 recordsLinked to original sources

arXiv2026

Do Input-Level Defenses Transfer to Observation-Level Attacks on VideoLLMs?

Video Large Language Models (VideoLLMs) are increasingly deployed in safety-critical applications such as content moderation and video analytics. To process long videos efficiently, VideoLLMs rely on frame sampling, token compression, and modality fusion, which together form an observation pipeline that reduces the raw video to a compact internal representation. Recent observation-level attacks exploit this pipeline to prevent the model from perceiving harmful content, yet no defense has been explicitly designed for this threat. We introduce DefTEval, a controlled evaluation framework that systematically assesses whether input-level adversarial defenses, which operate on the pixel content of already-sampled frames, can mitigate observation-level attacks. Across five VideoLLMs, eleven representative defenses, and five attack types, we find that input-level defenses offer limited and inconsistent protection, with harmful detection rates frequently near zero. Critically, defenses fail even against attacks that embed harmful signals in every sampled frame, indicating that the bottleneck extends beyond sampling omission to the suppression of signals that do enter the model. Token compression discards localized features, and modality fusion systematically down-weights weakened visual signals. Furthermore, defense effectiveness is dominated by model architecture rather than by the defense method itself, and detection rates vary drastically across content categories, exposing structural weaknesses in temporal reasoning. These findings demonstrate that securing VideoLLMs requires system-level robustness mechanisms spanning sampling-aware coverage guarantees, token-level preservation of safety-relevant features, and modality-balanced fusion.

cs.CV↗

arXiv2026

Unlocking Multi-Component Bulk-Materials Molecular Dynamics with a Small-Footprint Machine Learning Interatomic Potential

Bulk materials, as opposed to nanomaterials, require molecular dynamics (MD) simulations on a large spatial scale (~10^9 atoms or more) to adequately capture their atomic-scale physical properties. Previously, the introduction of machine-learning interatomic potentials (MLIPs) has extended MD to this scale, but even single-component bulk systems require tens of thousands of GPUs on high-end supercomputers. However, multi-component bulk MD simulations remain barely achievable, as the HBM footprint of existing MLIPs - already substantial for single-component systems - grows explosively in multi-component scenarios. This paper proposes an MLIP with a small HBM footprint - less than 3% that of existing MLIPs - unlocking multi-component bulk MD using only hundreds of GPUs. This is achieved by first identifying feature vectors and intermediate tensors as the two primary contributors to HBM footprints in existing MLIPs. To address these two sources, the dimensionality of the feature vectors has been reduced by introducing physical and chemical knowledge, and intermediate tensors have been eliminated by aggressively fusing all kernels into a single mega-kernel. In evaluation, the proposed MLIP has used 144 NVIDIA A100 GPUs to perform MD simulations on a 6-component bulk system with 1.14x10^9 atoms, while previously such MD simulation spatial scale has been restricted to unary systems and typically achieved on high-end supercomputers equipped with tens of thousands of GPUs.

physics.comp-ph↗

arXiv2026

Caved or Convinced: Temporal Sampling Gates Claim Deference in Video Large Language Models

When asked which of two events came first, video large language models can fail in two opposite ways: cave to a false claim, or reject a true one. Prior video sycophancy work measures only the first and mitigates it by teaching the model to trust the user less, a fix known in text and image models to worsen the second. In video, both failures come from two causes the literature treats as one: availability, whether the sparse sampled frames contain the two events, and weighting, whether that evidence is trusted over the user. We separate them with two interventions that keep the claim fixed: a frame-preserving reorder that flips the claim's truth, and a sampling-offset shift that captures or misses both events at a fixed frame budget. When the events are missed, the two twins present identical frames, so each of the nine models we evaluate accepts a true and a false claim at the same rate, making Youden's $J=0$ by construction. Availability is necessary but not sufficient. Five of the nine read the order, yet four of those five still cave to the false claim, so their deference hits a weighting ceiling. Since trust cannot be calibrated over evidence that was never sampled, we propose a reversal test that cancels the model's order prior by scoring the sampled frames forward and reversed, then answers, resamples, or abstains without reading the claim. The test raises the order accuracy to 0.92-1.00 on the models that read the order and abstains rather than guesses on those that cannot.

cs.MM↗

arXiv2026

SpreadMark: Robust Image Watermarking via Spread-Spectrum Embedding

Invisible image watermarks are increasingly used for deepfake detection and provenance tracking, where they must survive not only incidental distortions but also deliberate removal. We revisit spread-spectrum embedding, a classical watermarking principle, inside a modern neural post-hoc watermarking architecture. Our starting point is a measurement: in existing encoder-decoder schemes each message bit occupies only a small fraction of the image, a shared contributing factor to their fragility, since removal then need only disturb the region a bit occupies. SpreadMark instead spreads each bit as a dense pseudo-random codeword over the whole image and recovers it by matched-filtering a learned cover-suppressed chip representation, with a parallel convolutional decoding path and sparsification-aware training. A conditional chip-space analysis shows that, under a codeword-independent perturbation model, dense spreading increases the budget required to disrupt matched-filter recovery. Evaluated on COCO and DIV2K against nine schemes, SpreadMark is the only evaluated method retaining high detection under both the regeneration and the latent-space sparsification settings we test, with competitive JPEG and additive-noise robustness. It keeps the embedded watermark imperceptible, maintaining high perceptual quality on both COCO and DIV2K.

cs.CR↗

arXiv2026

VideoSTF: Stress-Testing Output Repetition in Video Large Language Models

Video Large Language Models (VideoLLMs) have recently achieved strong performance in video understanding tasks. However, we identify a previously underexplored generation failure: severe output repetition, where models degenerate into self-reinforcing loops of repeated phrases or sentences. This failure mode is not captured by existing VideoLLM benchmarks, which focus primarily on task accuracy and factual correctness. We introduce VideoSTF, the first framework for systematically measuring and stress-testing output repetition in VideoLLMs. VideoSTF formalizes repetition using three complementary n-gram-based metrics and provides a standardized testbed of 10,000 diverse videos together with a library of controlled temporal transformations. Using VideoSTF, we conduct pervasive testing, temporal stress testing, and adversarial exploitation across 10 advanced VideoLLMs. We find that output repetition is widespread and, critically, highly sensitive to temporal perturbations of video inputs. Moreover, we show that simple temporal transformations can efficiently induce repetitive degeneration in a black-box setting, exposing output repetition as an exploitable security vulnerability. Our results reveal output repetition as a fundamental stability issue in modern VideoLLMs and motivate stability-aware evaluation for video-language systems. Our evaluation code and scripts are available at: https://github.com/yuxincao22/VideoSTF_benchmark.

cs.CV↗

arXiv2026

FAIRT2V: Training-Free Debiasing for Text-to-Video Diffusion Models

Text-to-video (T2V) diffusion models have achieved rapid progress, yet their demographic biases, particularly gender bias, remain largely unexplored. We present FairT2V, a training-free debiasing framework for text-to-video generation that mitigates encoder-induced bias without finetuning. We first analyze demographic bias in T2V models and show that it primarily originates from pretrained text encoders, which encode implicit gender associations even for neutral prompts. We quantify this effect with a gender-leaning score that correlates with bias in generated videos. Based on this insight, FairT2V mitigates demographic bias by neutralizing prompt embeddings via anchor-based spherical geodesic transformations while preserving semantics. To maintain temporal coherence, we apply debiasing only during early identity-forming steps through a dynamic denoising schedule. We further propose a video-level fairness evaluation protocol combining VideoLLM-based reasoning with human verification. Experiments on the modern T2V model Open-Sora show that FairT2V substantially reduces demographic bias across occupations with minimal impact on video quality.

cs.CV↗

arXiv2026

DeMark: A Query-Free Black-Box Attack on Deepfake Watermarking Defenses

The rapid proliferation of realistic deepfakes has raised urgent concerns over their misuse, motivating the use of defensive watermarks in synthetic images for reliable detection and provenance tracking. However, this defense paradigm assumes such watermarks are inherently resistant to removal. We challenge this assumption with DeMark, a query-free black-box attack framework that targets defensive image watermarking schemes for deepfakes. DeMark exploits latent-space vulnerabilities in encoder-decoder watermarking models through a compressive sensing based sparsification process, suppressing watermark signals while preserving perceptual and structural realism appropriate for deepfakes. Across eight state-of-the-art watermarking schemes, DeMark reduces watermark detection accuracy from 100% to 32.9% on average while maintaining natural visual quality, outperforming existing attacks. We further evaluate three defense strategies, including image super resolution, sparse watermarking, and adversarial training, and find them largely ineffective. These results demonstrate that current encoder decoder watermarking schemes remain vulnerable to latent-space manipulations, underscoring the need for more robust watermarking methods to safeguard against deepfakes.

cs.CR↗

arXiv2026

Robust CAPTCHA Using Audio Illusions in the Era of Large Language Models: from Evaluation to Advances

CAPTCHAs are widely used by websites to block bots and spam by presenting challenges that are easy for humans but difficult for automated programs to solve. To improve accessibility, audio CAPTCHAs are designed to complement visual ones. However, the robustness of audio CAPTCHAs against advanced Large Audio Language Models (LALMs) and Automatic Speech Recognition (ASR) models remains unclear. In this paper, we introduce AI-CAPTCHA, a unified framework that offers (i) an evaluation framework, ACEval, which includes advanced LALM- and ASR-based solvers, and (ii) a novel audio CAPTCHA approach, IllusionAudio, leveraging audio illusions. Through extensive evaluations of seven widely deployed audio CAPTCHAs, we show that most existing methods can be solved with high success rates by advanced LALMs and ASR models, exposing critical security weaknesses. To address these vulnerabilities, we design a new audio CAPTCHA approach, IllusionAudio, which exploits perceptual illusion cues rooted in human auditory mechanisms. Extensive experiments demonstrate that our method defeats all tested LALM- and ASR-based attacks while achieving a 100% human pass rate, significantly outperforming existing audio CAPTCHA methods.

cs.SD↗

arXiv2025

Poisoning Prompt-Guided Sampling in Video Large Language Models

Video Large Language Models (VideoLLMs) are increasingly deployed as automated moderators on user-generated video platforms, where a few unwatched seconds of harmful footage are enough to suppress a safety alert. Because encoding every frame is prohibitive, modern VideoLLMs rely on prompt-guided sampling (PGS), which scores frames against the user prompt and forwards only the top-ranked ones to the visual encoder. Uniform and semantic samplers are known to be defeated by simple frame replacement, whereas PGS, the most prompt-aware family, has escaped scrutiny, and its prompt awareness in fact repairs the omission failures that defeat the other two. We show that this repair is superficial, since PoisonVID, a transfer attack, poisons the sampler's ranking so that harmful clips are never surfaced, without access to target weights, gradients, or sampling internals. It optimizes one video-level perturbation under a relevance-suppression loss defined over a depiction set of paraphrased harmful descriptions written by a shadow VideoLLM and a general-purpose language model, which drives perturbed harmful frames out of the prompt-conditioned subspace that PGS reads. Samplers that never consult that score keep the frames they always kept, which locates the failure at selection rather than at the encoder. Across three PGS methods, six VideoLLMs, and six harmful categories, PoisonVID attains 84% to 97% average attack success over the 18 sampler and model pairs and survives seven defenses. Re-encoding at lower resolution on ingest gives back part of what was evicted and costs the attack 48 points, which bounds the threat without closing it. PGS therefore buys accuracy with a structural safety debt, and sampler design will now have to repay that.

cs.CV↗

arXiv2025

Help or Hurdle? Rethinking Model Context Protocol-Augmented Large Language Models

The Model Context Protocol (MCP) enables large language models (LLMs) to access external resources on demand. While commonly assumed to enhance performance, how LLMs actually leverage this capability remains poorly understood. We introduce MCPGAUGE, the first comprehensive evaluation framework for probing LLM-MCP interactions along four key dimensions: proactivity (self-initiated tool use), compliance (adherence to tool-use instructions), effectiveness (task performance post-integration), and overhead (computational cost incurred). MCPGAUGE comprises a 160-prompt suite and 25 datasets spanning knowledge comprehension, general reasoning, and code generation. Our large-scale evaluation, spanning six commercial LLMs, 30 MCP tool suites, and both one- and two-turn interaction settings, comprises around 20,000 API calls and over USD 6,000 in computational cost. This comprehensive study reveals four key findings that challenge prevailing assumptions about the effectiveness of MCP integration. These insights highlight critical limitations in current AI-tool integration and position MCPGAUGE as a principled benchmark for advancing controllable, tool-augmented LLMs.

cs.AI↗

arXiv2025

Failures to Surface Harmful Contents in Video Large Language Models

Video Large Language Models (VideoLLMs) are increasingly deployed on numerous critical applications, where users rely on auto-generated summaries while casually skimming the video stream. We show that this interaction hides a critical safety gap: if harmful content is embedded in a video, either as full-frame inserts or as small corner patches, state-of-the-art VideoLLMs rarely mention the harmful content in the output, despite its clear visibility to human viewers. A root-cause analysis reveals three compounding design flaws: (1) insufficient temporal coverage resulting from the sparse, uniformly spaced frame sampling used by most leading VideoLLMs, (2) spatial information loss introduced by aggressive token downsampling within sampled frames, and (3) encoder-decoder disconnection, whereby visual cues are only weakly utilized during text generation. Leveraging these insights, we craft three zero-query black-box attacks, aligning with these flaws in the processing pipeline. Our large-scale evaluation across five leading VideoLLMs shows that the harmfulness omission rate exceeds 90% in most cases. Even when harmful content is clearly present in all frames, these models consistently fail to identify it. These results underscore a fundamental vulnerability in current VideoLLMs' designs and highlight the urgent need for sampling strategies, token compression, and decoding mechanisms that guarantee semantic coverage rather than speed alone.

cs.MM↗

arXiv2025

Accelerating Delta Debugging through Probabilistic Monotonicity Assessment

Delta debugging assumes search space monotonicity: if a program causes a failure, any supersets of that program will also induce the same failure, permitting the exclusion of subsets of non-failure-inducing programs. However, this assumption does not always hold in practice. This paper introduces Probabilistic Monotonicity Assessment (PMA), enhancing the efficiency of DDMIN-style algorithms without sacrificing effectiveness. PMA dynamically models and assesses the search space's monotonicity based on prior tests tried during the debugging process and uses a confidence function to quantify monotonicity, thereby enabling the probabilistic exclusion of subsets of non-failure-inducing programs. Our approach significantly reduces redundant tests that would otherwise be performed, without compromising the quality of the reduction. We evaluated PMA against two leading DDMIN-style tools, CHISEL and ProbDD. Our findings indicate that PMA cuts processing time by 59.2% compared to CHISEL, accelerates the reduction process (i.e., the number of tokens deleted per second) by 3.32x, and decreases the sizes of the final reduced programs by 6.7%. Against ProbDD, PMA reduces processing time by 22.0%, achieves a 1.34x speedup in the reduction process, and further decreases the sizes of the final reduced programs by 3.0%. These findings affirm PMA's role in significantly improving delta debugging's efficiency while maintaining or enhancing its effectiveness.

cs.SE↗

arXiv2025

ai.txt: A Domain-Specific Language for Guiding AI Interactions with the Internet

We introduce ai.txt, a novel domain-specific language (DSL) designed to explicitly regulate interactions between AI models, agents, and web content, addressing critical limitations of the widely adopted robots.txt standard. As AI increasingly engages with online materials for tasks such as training, summarization, and content modification, existing regulatory methods lack the necessary granularity and semantic expressiveness to ensure ethical and legal compliance. ai.txt extends traditional URL-based access controls by enabling precise element-level regulations and incorporating natural language instructions interpretable by AI systems. To facilitate practical deployment, we provide an integrated development environment with code autocompletion and automatic XML generation. Furthermore, we propose two compliance mechanisms: XML-based programmatic enforcement and natural language prompt integration, and demonstrate their effectiveness through preliminary experiments and case studies. Our approach aims to aid the governance of AI-Internet interactions, promoting responsible AI use in digital ecosystems.

cs.NI↗

arXiv2025

CFP: Efficient Optimization of Intra-Operator Parallelism Plans for Large Model Training

Optimizing the parallel training of large models requires exploring intra-operator parallelism plans for a computation graph that typically contains tens of thousands of primitive operators. While the optimization of parallel data processing graphs has been extensively researched in database systems, the vast search space makes it challenging to apply traditional database query optimization methods and algorithms. This paper introduces CFP, an optimization system for intra-operator parallelism that significantly reduces the complexity of searching for parallelism plans by leveraging two structural patterns found in large models. First, we identify parallel-preserving subgraphs, which ensure that the optimal global plan assigns the same parallel strategy to all operators within the subgraph. This approach allows us to avoid enumerating all possible combinations of parallel strategies for these operators. Second, we recognize repetitive subgraph patterns within the large computational graph, enabling us to profile a moderate number of representative subgraphs and accurately estimate the cost of parallelism plans with low overhead. With the significantly reduced search space, we can employ dynamic programming to search for the optimized parallelism plan. In our experiments, we demonstrate that CFP achieves significant speedups compared to the state-of-the-art framework for large models like GPT and LLAMA.

cs.DC↗

arXiv2022

Hybrid Inlining: A Compositional and Context Sensitive Static Analysis Framework

Context sensitivity is essential for achieving the precision in inter-procedural static analysis. To be (fully) context sensitive, top-down analysis needs to fully inline all statements of the callees at each callsite, leading to statement explosion. Compositional analysis, which inlines summaries of the callees, scales up but often loses precision, as it is not strictly context sensitive. We propose a compositional and strictly context sensitive framework for static analysis. This framework is based on one key observation: a compositional static analysis often loses precision only on some critical statements that need to be analyzed context sensitively. Our approach hybridly inlines the critical statements and the summaries of non-critical statements of each callee, thus avoiding the re-analysis of non-critical ones. In addition, our analysis lazily summarizes the critical statements, by stopping propagating the critical statements once the calling context accumulated is adequate. Hybrid Inlining can be as precise as context sensitive top-down analysis. We have designed and implemented a pointer analysis based on this framework. It can analyze large Java programs from the Dacapo benchmark suite and industry in minutes. In our evaluation, compared to context insensitive analysis, Hybrid Inlining just brings 65% and 1% additional time overhead on Dacapo and industrial applications respectively.

cs.LO↗

arXiv2020

Fusion-Catalyzed Pruning for Optimizing Deep Learning on Intelligent Edge Devices

The increasing computational cost of deep neural network models limits the applicability of intelligent applications on resource-constrained edge devices. While a number of neural network pruning methods have been proposed to compress the models, prevailing approaches focus only on parametric operators (e.g., convolution), which may miss optimization opportunities. In this paper, we present a novel fusion-catalyzed pruning approach, called FuPruner, which simultaneously optimizes the parametric and non-parametric operators for accelerating neural networks. We introduce an aggressive fusion method to equivalently transform a model, which extends the optimization space of pruning and enables non-parametric operators to be pruned in a similar manner as parametric operators, and a dynamic filter pruning method is applied to decrease the computational cost of models while retaining the accuracy requirement. Moreover, FuPruner provides configurable optimization options for controlling fusion and pruning, allowing much more flexible performance-accuracy trade-offs to be made. Evaluation with state-of-the-art residual neural networks on five representative intelligent edge platforms, Jetson TX2, Jetson Nano, Edge TPU, NCS, and NCS2, demonstrates the effectiveness of our approach, which can accelerate the inference of models on CIFAR-10 and ImageNet datasets.

cs.NE↗

arXiv2019

A Feature-Oriented Corpus for Understanding, Evaluating and Improving Fuzz Testing

Fuzzing is a promising technique for detecting security vulnerabilities. Newly developed fuzzers are typically evaluated in terms of the number of bugs found on vulnerable programs/binaries. However,existing corpora usually do not capture the features that prevent fuzzers from finding bugs, leading to ambiguous conclusions on the pros and cons of the fuzzers evaluated. A typical example is that Driller detects more bugs than AFL, but its evaluation cannot establish if the advancement of Driller stems from the concolic execution or not, since, for example, its ability in resolving a dataset`s magic values is unclear. In this paper, we propose to address the above problem by generating corpora based on search-hampering features. As a proof-of-concept, we have designed FEData, a prototype corpus that currently focuses on four search-hampering features to generate vulnerable programs for fuzz testing. Unlike existing corpora that can only answer "how", FEData can also further answer "why" by exposing (or understanding) the reasons for the identified weaknesses in a fuzzer. The "why" information serves as the key to the improvement of fuzzers.

cs.SE↗

arXiv2017

Understanding and Analyzing Java Reflection

Java reflection has been increasingly used in a wide range of software. It allows a software system to inspect and/or modify the behaviour of its classes, interfaces, methods and fields at runtime, enabling the software to adapt to dynamically changing runtime environments. However, this dynamic language feature imposes significant challenges to static analysis, because the behaviour of reflection-rich software is logically complex and statically hard to predict, especially when manipulated frequently by statically unknown string values. As a result, existing static analysis tools either ignore reflection or handle it partially, resulting in missed, important behaviours, i.e., unsound results. Therefore, improving or even achieving soundness in (static) reflection analysis -- an analysis that infers statically the behaviour of reflective code -- will provide significant benefits to many analysis clients, such as bug detectors, security analyzers and program verifiers. This paper makes two contributions: we provide a comprehensive understanding of Java reflection through examining its underlying concept, API and real-world usage, and, building on this, we introduce a new static approach to resolving Java reflection effectively in practice. We have implemented our reflection analysis in an open-source tool, called SOLAR, and evaluated its effectiveness extensively with large Java programs and libraries. Our experimental results demonstrate that SOLAR is able to (1) resolve reflection more soundly than the state-of-the-art reflection analysis; (2) automatically and accurately identify the parts of the program where reflection is resolved unsoundly or imprecisely; and (3) guide users to iteratively refine the analysis results by using lightweight annotations until their specific requirements are satisfied.

cs.PL↗