Searcharxiv⌕ Search

arXiv subjects

Shangzhi Xu

Publications and source records attributed to Shangzhi Xu.

9 recordsLinked to original sources

Growing Clusters Lower the Bar: Profile Accretion in LLM-Based Incremental Entity Resolution

Incremental entity resolution pipelines ask a large language model (LLM) whether a new record belongs to an existing cluster and show it the records already linked. Analyses of matching oracles assume that the answer depends only on the records compared. For the LLM matchers we test, it depends on cluster size, and the model lowers its bar for joining a cluster as the cluster grows. On French procurement suppliers from FOPPA, with SIREN identifiers as ground truth, we hold the new record fixed and vary only the number of verified members shown. Sixteen members instead of one raise the false match rate on similarly named companies by \dfmrMin--\dfmrMax{} percentage points in six open-weight models, while their area under the ROC curve rises by only \aucGainMin--\aucGainMax. In Claude Sonnet~5 and Haiku~5.5 the rise is \clSonD{} and \clHaiD{} points. The shift persists with street addresses, prompt variants and the BatchER and ComEM formats, and one false merge raises acceptance of the intruder's other records in every open model. It does not arise on Czech and Bulgarian TED notices, where one record with its full address already identifies the company. In online linking on confusable blocks, a single threshold set on labelled data raises recall by a factor of \fxRRmin--\fxRRmax{} for five models with no detectable loss of precision and no extra LLM calls.

cs.CL↗

CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems

Agent Skills allow LLM agents to combine reusable capabilities, but checking each Skill in isolation cannot rule out policy violations when they are composed. We term such failures Conflict of Skills (CoS): artifacts or assessment signals produced for one purpose may be reused downstream without preserving the restrictions or authorization needed for their new use. Using controlled executions of publicly available Skills and matched single-Skill controls, we identify four recurring mechanisms behind CoS. Building on these findings, we introduce the Conflict of Skills Attack (CoSA), which adds locally admitted Skills to an existing pool to supply missing business operations (capability supplementation) or reshape artifact exchange, decision reuse, and target/action bindings (relationship reshaping). CoSA leaves existing Skills and runtime permissions unchanged. To evaluate this threat under autonomous Skill selection, we introduce CoSA-Bench, built from real-world Skills and controlled services, with 288 tasks across recruiting, finance, IT, and business, covering data leakage, unauthorized state modification, and unsafe execution. Across five LLMs and three agent harnesses, additions capped at 5% of each original pool raise the policy-violation rate from 35.02% to 58.45%, an increase of 23.43 percentage points. Trace analysis links compliant-to-violating transitions to cross-Skill dependencies involving the additions. Finally, we develop a lightweight harness-level guard that checks information flow and authorization before consequential actions. On OpenClaw recruiting tasks across five models, the guard reduces violations from 65.28% to 6.67%, while policy-compliant task completion on matched benign controls changes from 66.67% to 65.00%.

cs.CL↗

Evaluating DeepSeek V4.1 Flash-1017 for loss of control and harmful manipulation risks

As frontier language models grow increasingly capable and autonomous, independent third-party evaluations are critical for characterizing their systemic safety risks. While closed-source releases are routinely accompanied by comprehensive system cards, open-weight models receive less scrutiny regarding frontier risks. To address this gap, we present an independent safety evaluation of DeepSeek V4.1 Flash-1017, the latest version of V4.1 Flash that departs from V4 with a new architecture, scaled-up training, and strong agentic capabilities. Drawing on the EU General-Purpose AI Code of Practice, we evaluate the model across 34 benchmarks, including 5 in-house suites, spanning evaluation integrity, loss of control, and harmful manipulation. Conforming to the AEF-1 standard, we compare V4.1 Flash-1017 against V4.1 Flash, earlier DeepSeek V4 models, and leading closed-source systems (GPT-6 Astra, GPT-5.6 Sol, and Gemini 3.8 Flash). Overall, V4.1 Flash-1017 shows substantial safety improvements over the V4 generation and approaches or matches leading closed-source baselines on several dimensions, while still exhibiting notable behaviors that warrant continued attention. Building on our findings, we highlight three key observations and recommendations. First, because safety can shift and even regress across minor updates, developers should integrate safety evaluation into incremental releases. Second, as final-turn outputs and end-state checks alone cannot reliably detect loss-of-control behaviors, deployments require process-level observability across both reasoning traces and action logs. Third, because public benchmarks elicit strong evaluation awareness and face data contamination risks, developing private benchmarks is essential for trustworthy frontier evaluation.

cs.CL↗

One Skill Too Many: How Co-Installed Skills Conflict in Coding Agents

Coding agents are extended with agent skills, directories whose SKILL.md tells the model when and how to perform a task. Because skills come from independent sources (teams, developers, plugins, copied collections), an installed skill can be co-installed with a similar skill doing the same job, and the model picks between them by name and description alone. In a conflict, the installed skill loses core functions (e.g., a ban on touching git) because the similar skill runs instead or changes what it does. The task still passes, so benchmarks that check only task completion miss such cases. We present the first empirical study of such conflicts. From snapshots of 20,947 repositories, we mine 822,109 candidate similar-skill pairs, have an LLM judge a stratified sample of 3,754, and run 312 confirmed pairs on three models (6,368 runs, 169,294 tool calls, 542 agent-hours). We report five findings. (1) Conflict-prone skills are common: nearly one in four installed skills is co-installed with one that does the same job, and 37% of judged skills sit inside copied collections. (2) Most such pairs involve normative skills, then capability skills. (3) Without lowering task completion, a similar skill takes one in five runs from the installed skill, and runs that open the similar skill first lose over a third of the exclusive core functions that only the installed skill fulfills. (4) Install location decides which skill runs, listing order barely matters, and the final reply names the skill used in only 0.9% of substituted runs. (5) Conflicts are decided at the first skill read, almost always before any file is changed, and a pre-tool hook at that read restores fidelity on exclusive core functions to the level of runs that open the installed skill first. Benchmarks should thus score exclusive core functions, and platforms should guard the first read and show which skill ran.

cs.SE↗

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

Video Large Language Models (VideoLLMs) have achieved strong performance on video understanding tasks, yet existing benchmarks evaluate only what models predict, leaving the stability of how they generate largely unexamined. We surface a previously underexplored generation failure of VideoLLMs, defined as output repetition, in which the decoder collapses into self-reinforcing loops of repeated phrases or sentences, and present VideoSTF, a benchmarking framework for systematically measuring, stress-testing, and exploiting this failure mode. VideoSTF formalizes repetition with three complementary $n$-gram-based metrics, ships a standardized testbed of 10,000 diverse videos, and provides a library of controlled temporal stressors. Across 10 advanced VideoLLMs, VideoSTF reveals four key findings: (i) repetition is pervasive on unperturbed videos and stable across commonly used frame counts, with repetition rates up to 91%; (ii) it spans a severity spectrum from mild redundancy to token-cap loops, and is highly amplified by temporal perturbations; (iii) temporal stressors form a practical black-box attack surface, flipping benign videos into repetitive ones with tens of queries and high attack success rates (up to 98%), and (iv) repetition is not explained by visual redundancy, its amplification tracks local temporal disruption, and only repetition penalties reduce it among common mitigations such as top-$k$ sampling, input filtering, and prompt variation, but increasing the penalty weakens visual grounding. VideoSTF reframes generation stability as a useful and complementary evaluation axis for VideoLLMs and provides the tools to study it. The project page is available at https://videostf.github.io/.

cs.CV↗

PUFFERDOS: Efficient and Effective Attack String Generation for Regular Expression Denial of Service Vulnerabilities

ReDoS attacks constitute a critical class of resource-exhaustion vulnerabilities. In such attacks, adversaries exploit the pathological worst-case execution behavior of regular expression (regex) engines to induce highly asymmetric computational workloads, ultimately exhausting system resources and degrading service availability. To protect systems against ReDoS attacks, numerous detection techniques have been proposed that simulate the attack process by generating attack strings to proactively exploit ReDoS vulnerabilities at the early development stage and facilitate remediation. Existing techniques broadly fall into two classes: static analyses that search for pathological regex structures, and dynamic exploration methods that synthesize candidate attack strings. However, the generated attack strings are often impractical for real-world exploitation because they usually assume unrealistic input-length budgets and do not validate the effectiveness and efficiency of the attack at the program level. Therefore, many generated strings fail to trigger vulnerable regexes when applied to real-world programs, further limiting the practical utility. To address these shortcomings, we introduce an effective and efficient attack string generator, PUFFERDOS, designed to synthesize attack inputs that are both feasible within realistic length budgets and validated at the program level, enabling effective exploitation of ReDoS vulnerabilities in real-world programs. Specifically, we first define three vulnerable patterns based on our observation and formal verification. According to the patterns, PUFFERDOS conducts a synthesis technique to generate attack strings, and then refines and validates the strings with ReDoS-specific compositional concolic execution to guarantee real-world exploitability.

cs.CR↗

A Globally Convergent Variational Framework for Mode Number Detection via Spectral Cutting Curves

Automatically determining the number of intrinsic mode functions (IMFs) and their center frequencies in Variational Mode Decomposition (VMD) remains an open mathematical challenge. Existing methods rely on heuristic settings, trial-and-error, or recursive extraction lacking theoretical convergence guarantees. We propose a variational framework that endogenously determines the number of modes. Any curve below the spectral amplitude divides the area under the spectrum into 2 parts and generate the connected intervals where spectrum locates above it, whose count defines the modal number K[g] -- a topological functional induced by the cutting curve. Since K[g] is discontinuous and intractable for direct optimization, we seek the optimal cutting curve as a continuous variational surrogate: it separates distinct spectral peaks into individual regions above it while merging noise-induced fragments below. This surrogate adversarially maximizes the integral of g while penalizing its curvature, transforming the problem into iteratively solving a fourth-order boundary value problem via Lagrangian duality. We establish a rigorous proof of global convergence for the dual ascent algorithm in function space. Comprehensive numerical experiments on artificial and real-world signals including ECG data show accurate estimates of IMFs and center frequencies, avoiding redundant modes while ensuring recovery of necessary components, providing a robust, theoretically grounded initialization routine for VMD.

math-ph↗

NGCaptcha: A CAPTCHA Bridging the Past and the Future

CAPTCHAs are widely employed for distinguishing humans from automated bots online. However, current vision based CAPTCHAs face escalating security risks: traditional attacks continue to bypass many deployed CAPTCHA schemes, and recent breakthroughs in AI, particularly large scale vision models, enable machine solvers to significantly outperform humans on many CAPTCHA tasks, undermining their original design assumptions. To address these issues, we introduce NGCAPTCHA, a Next Generation CAPTCHA framework that integrates a lightweight client side proof of work (PoW) mechanism with an AI resistant visual recognition challenge. In NGCAPTCHA, a browser must first complete a small hash based PoW before any challenge is displayed, throttling large scale automated attempts by increasing their computational cost. Once the PoW is solved, the user is presented with a human friendly yet model resistant image selection task that exploits perceptual cues current vision systems still struggle with. This hybrid design combines computational friction with AI robust visual discrimination, substantially raising the barrier for automated bots while keeping the verification process fast and effortless for legitimate users.

cs.CY↗

Enhancing Security in Third-Party Library Reuse -- Comprehensive Detection of 1-day Vulnerability through Code Patch Analysis

Nowadays, software development progresses rapidly to incorporate new features. To facilitate such growth and provide convenience for developers when creating and updating software, reusing open-source software (i.e., thirdparty library reuses) has become one of the most effective and efficient methods. Unfortunately, the practice of reusing third-party libraries (TPLs) can also introduce vulnerabilities (known as 1-day vulnerabilities) because of the low maintenance of TPLs, resulting in many vulnerable versions remaining in use. If the software incorporating these TPLs fails to detect the introduced vulnerabilities and leads to delayed updates, it will exacerbate the security risks. However, the complicated code dependencies and flexibility of TPL reuses make the detection of 1-day vulnerability a challenging task. To support developers in securely reusing TPLs during software development, we design and implement VULTURE, an effective and efficient detection tool, aiming at identifying 1-day vulnerabilities that arise from the reuse of vulnerable TPLs. It first executes a database creation method, TPLFILTER, which leverages the Large Language Model (LLM) to automatically build a unique database for the targeted platform. Instead of relying on code-level similarity comparison, VULTURE employs hashing-based comparison to explore the dependencies among the collected TPLs and identify the similarities between the TPLs and the target projects. Recognizing that developers have the flexibility to reuse TPLs exactly or in a custom manner, VULTURE separately conducts version-based comparison and chunk-based analysis to capture fine-grained semantic features at the function levels. We applied VULTURE to 10 real-world projects to assess its effectiveness and efficiency in detecting 1-day vulnerabilities. VULTURE successfully identified 175 vulnerabilities from 178 reused TPLs.

cs.SE↗