arXiv subjects

Meikang Qiu

Publications and source records attributed to Meikang Qiu.

At least 19 recordsLinked to original sources

arXiv2026

Can AI Agents Detect and Repair Artifact Drift in Network Experiments?

In recent years, AI agents have evolved into capable assistants that carry out multi-step tasks in digital environments. The network systems community is beginning to explore these capabilities in operational and experimental settings. However, an agent operating in network systems should not be judged solely by whether it completes the immediate task. The experiment record it modifies must also remain trustworthy. We call this property artifact integrity: the record's claims must remain supported by the available evidence, confined to the scope established by that evidence, and traceable through the artifacts that encode their support. To make this property measurable, we introduce NetArtifactBench, which tests whether AI agents can repair inconsistent records derived from public network-system artifacts while preserving claims that remain supported. The benchmark contains 52 instances with injected inconsistencies ranging from direct contradictions to unstated relations spread across several artifacts. We evaluate 23 agent configurations across three general-purpose AI agent runtimes using deterministic scoring. The average contract pass rate is 65.3 % across 5,980 outputs, but no agent runtime exceeds 30 % when repair requires recovering implicit relations and propagating changes across artifacts. These results reveal a sharp boundary between local correction and complete record-level repair. Therefore, we argue that artifact integrity should become a first-class design and evaluation requirement for AI agents operating on network systems.

cs.NI↗

arXiv2026

Can AI Agents Deliver Verifiable Network-Wide Outcomes Across Authority Boundaries?

AI agents are increasingly involved in network automation, where they can initiate configuration changes through mediated operational interfaces and assess the resulting state. Nonetheless, operational networks usually span many devices and administrative domains. Realizing an operator's intent requires coordinating agents with distinct authority scopes that define the resources they can access, the operations they can invoke, and the network state they can observe. This division limits the blast radius of an erroneous action but fragments the evidence needed to assess the network-wide outcome. Successful execution of a configuration action proposed by one agent does not establish that remote devices responded as intended or that routing changes reached the required devices. A valid observation may also become stale after a subsequent change. Before the coordinated operation can be declared complete, a trusted assurance layer must collect current observations from the required scopes and determine whether they collectively support the operator's intended network-wide outcome. To address the completion admission problem, we present EvidenceNet, a runtime assurance layer for deciding whether coordinated agent operations have achieved an operator's network intent. Its broker collects the post-change observations required by a completion contract, and its admission gate checks that the evidence comes from the required scopes, remains current, and satisfies the task rules. A verifier agent provides an additional assessment of the observation content. Experiments on live routing networks show that post-change state checks recognize successful outcomes that configuration-action records alone cannot establish. Controlled interventions further show that EvidenceNet rejects completion when otherwise satisfactory observations have the wrong source, have been substituted, or are stale.

cs.NI↗

arXiv2026

From Network Automation to Trustworthy Autonomous Networking in the LLM Era: A Network Control Intelligence Perspective

Since the inception of modern communication networks, the quest for operations automation has never ceased. Yet the evolution of network automation is difficult to characterize with a single maturity ladder. Throughout this history, network control systems have expanded their capabilities for observation, decision support, routine execution, and operator interaction, but these capabilities have not advanced uniformly. Such uneven progress makes the degree of automation an unreliable proxy for trustworthy network-side actuation. The unresolved question is not simply how much automation a system provides, but under what conditions it can be entrusted to change the network state. This paper examines that question through Network Control Intelligence (NCI), a five-axis framework spanning Decision Logic, Adaptability, Knowledge, Control Delegation, and Interface. We use NCI to organize the evolution of network-control systems into three eras: rule-based and scripted automation, programmable and data-driven control, and Large Language Model (LLM)-enabled network operations. Viewed through this framework, the three eras reveal a persistent asymmetry. None of these gains, however, automatically determines when network control should be trusted to change the network state. We frame trustworthy autonomy as a governed alignment between what a system can infer, what it can verify, and what it is authorized to execute. On that basis, the paper develops a reference architecture that separates proposal generation from governed execution, identifies recurring integration patterns for LLM-enabled operations, and derives a research agenda for higher network autonomy under explicit assurance, safety, and governance constraints.

cs.NI↗

arXiv2026

Beyond Refusal: A Same-Lineage Study of Aligned and Abliterated LLMs for Vulnerability Analysis

Large language model (LLM)-assisted software security operates at a difficult boundary: the vulnerability-analysis terminology needed for legitimate code review, triage, and repair can closely resemble terminology associated with misuse. Existing safety and cybersecurity evaluations are difficult to interpret in this setting because they often compare unrelated model families, thereby conflating safety behavior with differences in architecture, scale, training data, and deployment. To isolate this factor, we study safety state: whether refusal behavior remains intact (Aligned) or has been refusal-ablated (Abliterated) within same-lineage models. We ask how this safety state affects defensive utility across software-security workflows. We compare aligned instruction-tuned models with publicly released refusal-ablated descendants from two model families, Gemma and Qwen. We evaluate Aligned and Abliterated states on vulnerability detection, CWE attribution, vulnerable-line localization, root-cause localization, and executable patch validation. We further treat prompt wording as a controlled framing dimension: prompts begin with neutral code-review language, add authorization context, and vary the density of cybersecurity terminology. In a Gemma-based Java/Vul4J repair-validation study, Abliterated achieves higher early-stage validation rates, with 67.8%, 65.0%, and 32.8% of patches judged usable, successfully applied, and successfully compiled, respectively, compared with 29.9%, 24.9%, and 9.0% for Aligned. In the Qwen pair, Abliterated improves localization performance, increasing line-level F1 from 2.08% to 3.91% and Top-1 accuracy from 4.10% to 6.95%. These findings suggest that evaluations of LLM-based security assistants should jointly measure whether models respond, whether their usable responses are correct, and whether their outputs remain actionable across the engineering workflow.

cs.SE↗

arXiv2026

Image Prompt Reconstruction Attacks on Distributed MLLM Inference Frameworks

Distributed large language model (LLM) inference frameworks connect isolated consumer-grade devices for large-scale model inference, substantially reducing hardware constraints. However, recent studies show that intermediate embeddings transmitted among participants can leak private prompts. As LLMs evolve into multimodal LLMs (MLLMs), this risk extends beyond text: image prompts contain rich visual and semantic information, making their intermediate embeddings highly privacy-sensitive. Yet, image-prompt leakage in distributed MLLM inference remains largely unexplored. In this paper, we investigate privacy risks to input images caused by intermediate embeddings in distributed MLLM frameworks. We first analyze the information flow from image pixels to intermediate representations. Since image and text embeddings are often intertwined across MLLM layers, we design an image embedding extraction algorithm as a prerequisite for reconstruction attacks, achieving 100% extraction accuracy across almost all MLLM layers in our experiments. Building on this, we develop two passive black-box image reconstruction attacks, MPAA and IEDA, reflecting realistic threats from normal participants with limited knowledge and capability. MPAA performs fine-grained pixel-level reconstruction via patch-wise information extraction and assembly, while IEDA performs coarse-grained semantic reconstruction through embedding-guided diffusion generation. We evaluate our attacks on four representative MLLM families: Gemma 3, Phi 4 Multimodal, Qwen 2.5 VL, and Llama 4 Scout. Results show consistently superior reconstruction performance in various settings. We further analyze the effects of MoE architecture, image preprocessing, model size, and text-image dependency on attack performance. To our knowledge, this is the first study of image reconstruction attacks on MLLMs.

cs.CR↗

arXiv2025

ScamSweeper: Detecting Illegal Accounts in Web3 Scams via Transactions Analysis

The web3 applications have recently been growing, especially on the Ethereum platform, starting to become the target of scammers. The web3 scams, imitating the services provided by legitimate platforms, mimic regular activity to deceive users. However, previous studies have primarily concentrated on de-anonymization and phishing nodes, neglecting the distinctive features of web3 scams. Moreover, the current phishing account detection tools utilize graph learning or sampling algorithms to obtain graph features. However, large-scale transaction networks with temporal attributes conform to a power-law distribution, posing challenges in detecting web3 scams. To overcome these challenges, we present ScamSweeper, a novel framework that emphasizes the dynamic evolution of transaction graphs, to identify web3 scams on Ethereum. ScamSweeper samples the network with a structure temporal random walk, which is an optimized sample walking method that considers both temporal attributes and structural information. Then, the directed graph encoder generates the features of each subgraph during different temporal intervals, sorting as a sequence. Moreover, a variational Transformer is utilized to extract the dynamic evolution in the subgraph sequence. Furthermore, we collect a large-scale transaction dataset consisting of web3 scams, phishing, and normal accounts, which are from the first 18 million block heights on Ethereum. Subsequently, we comprehensively analyze the distinctions in various attributes, including nodes, edges, and degree distribution. Our experiments indicate that ScamSweeper outperforms SIEGE, Ethident, and PDTGA in detecting web3 scams, achieving a weighted F1-score improvement of at least 17.29% with the base value of 0.59. In addition, ScamSweeper in phishing node detection achieves at least a 17.5% improvement over DGTSG and BERT4ETH in F1-score from 0.80.

cs.CR↗

arXiv2025

MIMO: A medical vision language model with visual referring multimodal input and pixel grounding multimodal output

Currently, medical vision language models are widely used in medical vision question answering tasks. However, existing models are confronted with two issues: for input, the model only relies on text instructions and lacks direct understanding of visual clues in the image; for output, the model only gives text answers and lacks connection with key areas in the image. To address these issues, we propose a unified medical vision language model MIMO, with visual referring Multimodal Input and pixel grounding Multimodal Output. MIMO can not only combine visual clues and textual instructions to understand complex medical images and semantics, but can also ground medical terminologies in textual output within the image. To overcome the scarcity of relevant data in the medical field, we propose MIMOSeg, a comprehensive medical multimodal dataset including 895K samples. MIMOSeg is constructed from four different perspectives, covering basic instruction following and complex question answering with multimodal input and multimodal output. We conduct experiments on several downstream medical multimodal tasks. Extensive experimental results verify that MIMO can uniquely combine visual referring and pixel grounding capabilities, which are not available in previous models.

cs.CV↗

arXiv2025

DARD: Dice Adversarial Robustness Distillation against Adversarial Attacks

Deep learning models are vulnerable to adversarial examples, posing critical security challenges in real-world applications. While Adversarial Training (AT ) is a widely adopted defense mechanism to enhance robustness, it often incurs a trade-off by degrading performance on unperturbed, natural data. Recent efforts have highlighted that larger models exhibit enhanced robustness over their smaller counterparts. In this paper, we empirically demonstrate that such robustness can be systematically distilled from large teacher models into compact student models. To achieve better performance, we introduce Dice Adversarial Robustness Distillation (DARD), a novel method designed to transfer robustness through a tailored knowledge distillation paradigm. Additionally, we propose Dice Projected Gradient Descent (DPGD), an adversarial example generalization method optimized for effective attack. Our extensive experiments demonstrate that the DARD approach consistently outperforms adversarially trained networks with the same architecture, achieving superior robustness and standard accuracy.

cs.LG↗

arXiv2025

From Scores to Skills: A Cognitive Diagnosis Framework for Evaluating Financial Large Language Models

Large Language Models (LLMs) have shown promise for financial applications, yet their suitability for this high-stakes domain remains largely unproven due to inadequacies in existing benchmarks. Existing benchmarks solely rely on score-level evaluation, summarizing performance with a single score that obscures the nuanced understanding of what models truly know and their precise limitations. They also rely on datasets that cover only a narrow subset of financial concepts, while overlooking other essentials for real-world applications. To address these gaps, we introduce FinCDM, the first cognitive diagnosis evaluation framework tailored for financial LLMs, enabling the evaluation of LLMs at the knowledge-skill level, identifying what financial skills and knowledge they have or lack based on their response patterns across skill-tagged tasks, rather than a single aggregated number. We construct CPA-KQA, the first cognitively informed financial evaluation dataset derived from the Certified Public Accountant (CPA) examination, with comprehensive coverage of real-world accounting and financial skills. It is rigorously annotated by domain experts, who author, validate, and annotate questions with high inter-annotator agreement and fine-grained knowledge labels. Our extensive experiments on 30 proprietary, open-source, and domain-specific LLMs show that FinCDM reveals hidden knowledge gaps, identifies under-tested areas such as tax and regulatory reasoning overlooked by traditional benchmarks, and uncovers behavioral clusters among models. FinCDM introduces a new paradigm for financial LLM evaluation by enabling interpretable, skill-aware diagnosis that supports more trustworthy and targeted model development, and all datasets and evaluation scripts will be publicly released to support further research.

cs.CE↗

arXiv2025

AutoDebias: Automated Framework for Debiasing Text-to-Image Models

Text-to-Image (T2I) models generate high-quality images but are vulnerable to malicious backdoor attacks that inject harmful biases (e.g., trigger-activated gender or racial stereotypes). Existing debiasing methods, often designed for natural statistical biases, struggle with these deliberately and subtly injected attacks. We propose AutoDebias, a framework that automatically identifies and mitigates these malicious biases in T2I models without prior knowledge of the specific attack types. Specifically, AutoDebias leverages vision-language models to detect trigger-activated visual patterns and constructs neutralization guides by generating counter-prompts. These guides drive a CLIP-guided training process that breaks the harmful associations while preserving the original model's image quality and diversity. Unlike methods designed for natural bias, AutoDebias effectively addresses subtle, injected stereotypes and multiple interacting attacks. We evaluate the framework on a new benchmark covering 17 distinct backdoor scenarios, including challenging cases where multiple backdoors co-exist. AutoDebias detects malicious patterns with 91.6% accuracy and reduces the backdoor success rate from 90% to negligible levels, while preserving the visual fidelity of the original model.

cs.CV↗

arXiv2025

MultiFinBen: Benchmarking Large Language Models for Multilingual and Multimodal Financial Application

Real-world financial analysis involves information across multiple languages and modalities, from reports and news to scanned filings and meeting recordings. Yet most existing evaluations of LLMs in finance remain text-only, monolingual, and largely saturated by current models. To bridge these gaps, we present MultiFinBen, the first expert-annotated multilingual (five languages) and multimodal (text, vision, audio) benchmark for evaluating LLMs in realistic financial contexts. MultiFinBen introduces two new task families: multilingual financial reasoning, which tests cross-lingual evidence integration from filings and news, and financial OCR, which extracts structured text from scanned documents containing tables and charts. Rather than aggregating all available datasets, we apply a structured, difficulty-aware selection based on advanced model performance, ensuring balanced challenge and removing redundant tasks. Evaluating 21 leading LLMs shows that even frontier multimodal models like GPT-4o achieve only 46.01% overall, stronger on vision and audio but dropping sharply in multilingual settings. These findings expose persistent limitations in multilingual, multimodal, and expert-level financial reasoning. All datasets, evaluation scripts, and leaderboards are publicly released.

cs.CL↗

arXiv2025

PrunePEFT: Iterative Hybrid Pruning for Parameter-Efficient Fine-tuning of LLMs

Parameter Efficient Fine-Tuning (PEFT) methods have emerged as effective and promising approaches for fine-tuning pre-trained language models. Compared with Full parameter Fine-Tuning (FFT), PEFT achieved comparable task performance with a substantial reduction of trainable parameters, which largely saved the training and storage costs. However, using the PEFT method requires considering a vast design space, such as the type of PEFT modules and their insertion layers. Inadequate configurations can lead to sub-optimal results. Conventional solutions such as architectural search techniques, while effective, tend to introduce substantial additional overhead. In this paper, we propose a novel approach, PrunePEFT, which formulates the PEFT strategy search as a pruning problem and introduces a hybrid pruning strategy that capitalizes on the sensitivity of pruning methods to different PEFT modules. This method extends traditional pruning techniques by iteratively removing redundant or conflicting PEFT modules, thereby optimizing the fine-tuned configuration. By efficiently identifying the most relevant modules, our approach significantly reduces the computational burden typically associated with architectural search processes, making it a more scalable and efficient solution for fine-tuning large pre-trained models.

cs.LG↗

arXiv2025

Towards Effective Federated Graph Foundation Model via Mitigating Knowledge Entanglement

Recent advances in graph machine learning have shifted to data-centric paradigms, driven by two emerging fields: (1) Federated graph learning (FGL) enables multi-client collaboration but faces challenges from data and task heterogeneity, limiting its practicality; (2) Graph foundation models (GFM) offer strong domain generalization but are usually trained on single machines, missing out on cross-silo data and resources. These paradigms are complementary, and their integration brings notable benefits. Motivated by this, we propose FedGFM, a novel decentralized GFM training paradigm. However, a key challenge is knowledge entanglement, where multi-domain knowledge merges into indistinguishable representations, hindering downstream adaptation. To address this, we present FedGFM+, an enhanced framework with two core modules to reduce knowledge entanglement: (1) AncDAI: A global anchor-based domain-aware initialization strategy. Before pre-training, each client encodes its local graph into domain-specific prototypes that serve as semantic anchors. Synthetic embeddings around these anchors initialize the global model. We theoretically prove these prototypes are distinguishable across domains, providing a strong inductive bias to disentangle domain-specific knowledge. (2) AdaDPP: A local adaptive domain-sensitive prompt pool. Each client learns a lightweight graph prompt capturing domain semantics during pre-training. During fine-tuning, prompts from all clients form a pool from which the GFM selects relevant prompts to augment target graph attributes, improving downstream adaptation. FedGFM+ is evaluated on 8 diverse benchmarks across multiple domains and tasks, outperforming 20 baselines from supervised learning, FGL, and federated GFM variants.

cs.LG↗

arXiv2025

T2S: High-resolution Time Series Generation with Text-to-Series Diffusion Models

Text-to-Time Series generation holds significant potential to address challenges such as data sparsity, imbalance, and limited availability of multimodal time series datasets across domains. While diffusion models have achieved remarkable success in Text-to-X (e.g., vision and audio data) generation, their use in time series generation remains in its nascent stages. Existing approaches face two critical limitations: (1) the lack of systematic exploration of general-proposed time series captions, which are often domain-specific and struggle with generalization; and (2) the inability to generate time series of arbitrary lengths, limiting their applicability to real-world scenarios. In this work, we first categorize time series captions into three levels: point-level, fragment-level, and instance-level. Additionally, we introduce a new fragment-level dataset containing over 600,000 high-resolution time series-text pairs. Second, we propose Text-to-Series (T2S), a diffusion-based framework that bridges the gap between natural language and time series in a domain-agnostic manner. T2S employs a length-adaptive variational autoencoder to encode time series of varying lengths into consistent latent embeddings. On top of that, T2S effectively aligns textual representations with latent embeddings by utilizing Flow Matching and employing Diffusion Transformer as the denoiser. We train T2S in an interleaved paradigm across multiple lengths, allowing it to generate sequences of any desired length. Extensive evaluations demonstrate that T2S achieves state-of-the-art performance across 13 datasets spanning 12 domains.

cs.LG↗

arXiv2025

FinAudio: A Benchmark for Audio Large Language Models in Financial Applications

Audio Large Language Models (AudioLLMs) have received widespread attention and have significantly improved performance on audio tasks such as conversation, audio understanding, and automatic speech recognition (ASR). Despite these advancements, there is an absence of a benchmark for assessing AudioLLMs in financial scenarios, where audio data, such as earnings conference calls and CEO speeches, are crucial resources for financial analysis and investment decisions. In this paper, we introduce \textsc{FinAudio}, the first benchmark designed to evaluate the capacity of AudioLLMs in the financial domain. We first define three tasks based on the unique characteristics of the financial domain: 1) ASR for short financial audio, 2) ASR for long financial audio, and 3) summarization of long financial audio. Then, we curate two short and two long audio datasets, respectively, and develop a novel dataset for financial audio summarization, comprising the \textsc{FinAudio} benchmark. Then, we evaluate seven prevalent AudioLLMs on \textsc{FinAudio}. Our evaluation reveals the limitations of existing AudioLLMs in the financial domain and offers insights for improving AudioLLMs. All datasets and codes will be released.

cs.CE↗

arXiv2025

DR.GAP: Mitigating Bias in Large Language Models using Gender-Aware Prompting with Decoupled Reasoning

Large Language Models (LLMs) exhibit strong natural language understanding capabilities but also inherit and amplify societal biases, particularly gender bias, raising fairness concerns. Existing prompt-based debiasing strategies share a key limitation: they fail to disentangle gender information from task semantics. Bias steering compels models to overemphasize gender cues, while reasoning-based prompting induces gender-biased reasoning chains. To address these challenges, we propose DR.GAP (Decoupled Reasoning for Gender-Aware Prompting), an automated and model-agnostic pipeline that mitigates gender bias while preserving model performance. DR.GAP generates gender-neutral reasoning traces and applies them as in-context demonstrations during inference, effectively decoupling gender attributes from task semantics without modifying model parameters. Extensive experiments on coreference resolution and question-answering tasks across six LLMs demonstrate DR.GAP's effectiveness, generalizability, and robustness, supported by detailed mechanism analyses. Moreover, DR.GAP can be extended to vision-language models (VLMs), achieving substantial bias reduction.

cs.CL↗

arXiv2024

Open-FinLLMs: Open Multimodal Large Language Models for Financial Applications

Financial LLMs hold promise for advancing financial tasks and domain-specific applications. However, they are limited by scarce corpora, weak multimodal capabilities, and narrow evaluations, making them less suited for real-world application. To address this, we introduce \textit{Open-FinLLMs}, the first open-source multimodal financial LLMs designed to handle diverse tasks across text, tabular, time-series, and chart data, excelling in zero-shot, few-shot, and fine-tuning settings. The suite includes FinLLaMA, pre-trained on a comprehensive 52-billion-token corpus; FinLLaMA-Instruct, fine-tuned with 573K financial instructions; and FinLLaVA, enhanced with 1.43M multimodal tuning pairs for strong cross-modal reasoning. We comprehensively evaluate Open-FinLLMs across 14 financial tasks, 30 datasets, and 4 multimodal tasks in zero-shot, few-shot, and supervised fine-tuning settings, introducing two new multimodal evaluation datasets. Our results show that Open-FinLLMs outperforms afvanced financial and general LLMs such as GPT-4, across financial NLP, decision-making, and multi-modal tasks, highlighting their potential to tackle real-world challenges. To foster innovation and collaboration across academia and industry, we release all codes (https://anonymous.4open.science/r/PIXIU2-0D70/B1D7/LICENSE) and models under OSI-approved licenses.

cs.CL↗

arXiv2024

Self-consistent Deep Geometric Learning for Heterogeneous Multi-source Spatial Point Data Prediction

Multi-source spatial point data prediction is crucial in fields like environmental monitoring and natural resource management, where integrating data from various sensors is the key to achieving a holistic environmental understanding. Existing models in this area often fall short due to their domain-specific nature and lack a strategy for integrating information from various sources in the absence of ground truth labels. Key challenges include evaluating the quality of different data sources and modeling spatial relationships among them effectively. Addressing these issues, we introduce an innovative multi-source spatial point data prediction framework that adeptly aligns information from varied sources without relying on ground truth labels. A unique aspect of our method is the 'fidelity score,' a quantitative measure for evaluating the reliability of each data source. Furthermore, we develop a geo-location-aware graph neural network tailored to accurately depict spatial relationships between data points. Our framework has been rigorously tested on two real-world datasets and one synthetic dataset. The results consistently demonstrate its superior performance over existing state-of-the-art methods.

cs.LG↗