arXiv subjects

Ting Wang

Publications and source records attributed to Ting Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Development of a sub-100 ps Time-of-Flight detector with SiPM-readout scintillator for measurement of cosmic muon velocity

Accurate Time-of-Flight (TOF) measurement with sub-100 picosecond resolution is a critical requirement for particle identification in future high-energy physics experiments, such as the Belle II $K_{L}$ and Muon (KLM) detector upgrade. Achieving this precision with large-area Silicon Photomultipliers (SiPMs) is challenging due to the inherent junction capacitance, which degrades signal rise time. In this work, we developed and evaluated a high-time-resolution cosmic ray detector based on plastic scintillators and customized SiPM arrays. To optimize the readout for block-shaped scintillators, we systematically compared different sensor topologies. We demonstrate that a multi-face readout topology, utilizing low-capacitance 4-series (4S) SiPM modules coupled to four faces of the scintillator, achieves an excellent coincidence time resolution of approximately 68 ps, outperforming the $\sim$100 ps resolution of the concentrated 4-series 3-parallel (4S3P) hybrid topology. Furthermore, to validate the system's practical performance, we successfully measured well-known cosmic ray observables, specifically the relativistic muon velocity via TOF reconstruction. These results highlight the potential of the multi-face 4S configuration as a high-precision solution for future TOF detector upgrades.

physics.ins-det↗

arXiv2026

Hybrid Quantum-MambaVision: A Quantum-enhanced State Space Model for Calibrated Mixed-type Wafer Defect Detection

Extracting actionable knowledge from industrial visual data is fundamentally bottlenecked by extreme class imbalance and the prohibitive computational complexity of modern foundation models. In semi-conductor manufacturing, identifying multi-label wafer defects is a complex spatial data mining task where overlapping patterns obscure critical root-cause signals. While Vision Transformers (ViTs) excel at global dependency extraction, their quadratic scaling renders them inefficient for high-throughput, real-time anomaly detection. To overcome these computational barriers, this paper introduces Hybrid Quantum-MambaVision, a highly efficient architecture tailored for spatial knowledge discovery. We integrate a linear-complexity State-Space Model (SSM) backbone with a Parameterized Quantum Context Adapter (QCA) and Low-Rank Adaptation (LoRA). The Mamba backbone efficiently captures long-range spatial dependencies, while the quantum adapter maps compressed latent features into a high-dimensional Hilbert space to disentangle complex, overlapping signatures. On the highly imbalanced MixedWM38 dataset, Hybrid Quantum-MambaVision achieves exceptional multi-label classification performance, significantly reducing the error rate on complex multi-defect topologies compared to classical baselines. The quantum regularizer acts as a profound uncertainty calibrator, substantially reducing Maximum Calibration Error (MCE) and minimizing expected false-positive costs. This work establishes a scalable Quantum-Classical hybrid paradigm for efficient representation learning in industrial data mining.

cs.CV↗

arXiv2026

Generative diffusion learning for parametric partial differential equations

We develop a class of data-driven generative models that approximate the solution operator for parameter-dependent partial differential equations (PDE). We propose a novel probabilistic formulation of the operator learning problem based on recently developed generative denoising diffusion probabilistic models (DDPM) in order to learn the input-to-output mapping between problem parameters and solutions of the PDE. To achieve this goal we modify DDPM to supervised learning in which the solution operator for the PDE is represented by a class of conditional distributions. The probabilistic formulation combined with DDPM allows for an automatic quantification of confidence intervals for the learned solutions. Furthermore, the framework is directly applicable for learning from a noisy data set. We compare computational performance of the developed method with the Fourier Network Operators (FNO). Our results show that our method achieves comparable accuracy and recovers the noise magnitude when applied to data sets with outputs corrupted by additive noise.

math.NA↗

arXiv2026

AutoRAN: Automated Hijacking of Safety Reasoning in Large Reasoning Models

This paper presents AutoRAN, the first framework to automate the hijacking of internal safety reasoning in large reasoning models (LRMs). At its core, AutoRAN pioneers an execution simulation paradigm that leverages a weaker but less-aligned model to simulate execution reasoning for initial hijacking attempts and iteratively refine attacks by exploiting reasoning patterns leaked through the target LRM's refusals. This approach steers the target model to bypass its own safety guardrails and elaborate on harmful instructions. We evaluate AutoRAN against state-of-the-art LRMs, including GPT-o3/o4-mini and Gemini-2.5-Flash, across multiple benchmarks (AdvBench, HarmBench, and StrongReject). Results show that AutoRAN achieves approaching 100% success rate within one or few turns, effectively neutralizing reasoning-based defenses even when evaluated by robustly aligned external models. This work reveals that the transparency of the reasoning process itself creates a critical and exploitable attack surface, highlighting the urgent need for new defenses that protect models' reasoning traces rather than merely their final outputs.

cs.LG↗

arXiv2026

Spectral analysis of multivariate stationary Hawkes processes

We establish the asymptotic validity of frequency-domain inference for stationary multivariate Hawkes processes under mild conditions, bridging the gap between theory and application. By developing upper-bounds on the reduced cumulant measures from the cluster representation of the Hawkes processes, we prove a functional central limit theorem and, as a consequence, consistency of the Whittle estimator under stationarity alone (i.e., the spectral radius of the interactions matrix $ρ(\boldsymbolν)<1$), applicable to Hawkes processes with heavy-tailed mutual-excitation kernels. Under mild extra moment conditions, we further obtain asymptotic normality with an explicit limiting covariance in terms of second- and fourth-order cumulant spectral densities. We also propose a simple frequency-domain method to detect joint independence of subprocesses of a multivariate Hawkes process. The performance of the Whittle estimator and the test of independence are demonstrated via simulation studies.

math.ST↗

arXiv2026

ProVG: Progressive Visual Grounding via Language Decoupling for Remote Sensing Imagery

Remote sensing visual grounding (RSVG) aims to localize objects in remote sensing imagery according to natural language expressions. Previous methods typically rely on sentence-level vision-language alignment, which struggles to exploit fine-grained linguistic cues, such as \textit{spatial relations} and \textit{object attributes}, that are crucial for distinguishing objects with similar characteristics. Importantly, these cues play distinct roles across different grounding stages and should be leveraged accordingly to provide more explicit guidance. In this work, we propose \textbf{ProVG}, a novel RSVG framework that improves localization accuracy by decoupling language expressions into global context, spatial relations, and object attributes. To integrate these linguistic cues, ProVG employs a simple yet effective progressive cross-modal modulator, which dynamically modulates visual attention through a \textit{survey-locate-verify} scheme, enabling coarse-to-fine vision-language alignment. In addition, ProVG incorporates a cross-scale fusion module to mitigate the large-scale variations in remote sensing imagery, along with a language-guided calibration decoder to refine cross-modal alignment during prediction. A unified multi-task head further enables ProVG to support both referring expression comprehension and segmentation tasks. Extensive experiments on two benchmarks, \textit{i.e.}, RRSIS-D and RISBench, demonstrate that ProVG consistently outperforms existing methods, achieving new state-of-the-art performance.

cs.CV↗

arXiv2026

Tailoring Quasi-Bound States in the Continuum for Infrared Photodetection in Black Phosphorus

High-performance infrared photodetection underpins various applications spanning surveillance, environmental monitoring, optical communications and biomedical imaging. However, conventional bulk detectors remain limited by poor spectral tunability, mechanical rigidity, and high dark currents, motivating the pursuit of low-dimensional material platforms such as graphene and transition metal dichalgenides. Black phosphorus (BP) is particularly compelling in this context, owing to its thickness-tunable direct bandgap, high carrier mobility, and pronounced in-plane anisotropy. Nevertheless, its atomically thin nature inherently restricts light absorption, posing a fundamental bottleneck for device performance. Here, we demonstrate quasi-bound states in the continuum (quasi-BICs) within a dielectric metasurface integrated with BP, enabling strongly enhanced and spectrally selective light-matter interactions. By introducing controlled symmetry breaking at the unit-cell level, high-quality-factor resonances are realized, resulting in pronounced electromagnetic field confinement within the BP layer. This resonant enhancement substantially increases photocarrier generation while preserving the intrinsic polarization anisotropy of BP, which elucidates a robust pathway for overcoming the optical absorption bottleneck in anisotropic 2D optoelectronics via quasi-BIC platforms.

physics.optics↗

arXiv2026

MetaKube: An Experience-Aware LLM Framework for Kubernetes Failure Diagnosis

Existing LLM-based Kubernetes diagnostic systems cannot learn from operational experience, operating on static knowledge bases without improving from past resolutions. We present MetaKube, an experience-aware LLM framework through three synergistic innovations: (1) an Episodic Pattern Memory Network (EPMN) that abstracts diagnostic patterns from historical resolutions and provides confidence-calibrated retrieval for both rapid pattern matching and guided causal exploration, (2) a meta-cognitive controller that dynamically routes between intuitive and analytical pathways based on problem familiarity, optimizing the trade-off between speed and depth, and (3) KubeLLM, a locally-deployable 8B model enhanced through domain-specific post-training on our 7,000-sample Kubernetes Fault Resolution Dataset. Evaluation on 1,873 real-world scenarios demonstrates MetaKube transforms Qwen3-8B from 50.9 to 90.5 points, approaching GPT-4.1 performance while ensuring complete data privacy. EPMN contributes 15.3% improvement through experiential learning, with continuous learning experiments showing progressive gains as the system accumulates operational knowledge. The source code and related resources are available at https://github.com/MetaKube-LLM-for-Kubernetes-Diagnosis/MetaKube.

cs.LG↗

arXiv2026

Self-Destructive Language Model

Harmful fine-tuning attacks pose a major threat to the security of large language models (LLMs), allowing adversaries to compromise safety guardrails with minimal harmful data. While existing defenses attempt to reinforce LLM alignment, they fail to address models' inherent "trainability" on harmful data, leaving them vulnerable to stronger attacks with increased learning rates or larger harmful datasets. To overcome this critical limitation, we introduce SEAM, a novel alignment-enhancing defense that transforms LLMs into self-destructive models with intrinsic resilience to misalignment attempts. Specifically, these models retain their capabilities for legitimate tasks while exhibiting substantial performance degradation when fine-tuned on harmful data. The protection is achieved through a novel loss function that couples the optimization trajectories of benign and harmful data, enhanced with adversarial gradient ascent to amplify the self-destructive effect. To enable practical training, we develop an efficient Hessian-free gradient estimate with theoretical error bounds. Extensive evaluation across LLMs and datasets demonstrates that SEAM creates a no-win situation for adversaries: the self-destructive models achieve state-of-the-art robustness against low-intensity attacks and undergo catastrophic performance collapse under high-intensity attacks, rendering them effectively unusable. The code is available: https://github.com/ZJUWYH/seam. (Warning: this paper contains potentially harmful content generated by LLMs.)

cs.LG↗

arXiv2026

Reasoning or Retrieval? A Study of Answer Attribution on Large Reasoning Models

Large reasoning models (LRMs) exhibit unprecedented capabilities in solving complex problems through Chain-of-Thought (CoT) reasoning. However, recent studies reveal that their final answers often contradict their own reasoning traces. We hypothesize that this inconsistency stems from two competing mechanisms for generating answers: CoT reasoning and memory retrieval. To test this hypothesis, we conduct controlled experiments that challenge LRMs with misleading cues during reasoning and/or corrupted answers during retrieval. Our results across models and datasets confirm that both mechanisms operate simultaneously, with their relative dominance influenced by multiple factors: problem domains, model scales, and fine-tuning approaches (e.g., reinforcement learning vs. distillation). The findings reveal a critical limitation in current reasoning fine-tuning paradigms: models can exploit the retrieval mechanism as a shortcut, effectively "hacking" the reward signal and undermining genuine reasoning development. To address this challenge, we introduce FARL, a novel fine-tuning framework that integrates memory unlearning with reinforcement learning. By carefully suppressing retrieval shortcuts during the fine-tuning process, FARL promotes reasoning-dominant behavior and enhances generalizable reasoning capabilities. The code is available: https://github.com/ZJUWYH/FARL.

cs.AI↗

arXiv2026

Dynamic Token Reweighting for Robust Vision-Language Models

Large vision-language models (VLMs) are highly vulnerable to multimodal jailbreak attacks that exploit visual-textual interactions to bypass safety guardrails. In this paper, we present DTR, a novel inference-time defense that mitigates multimodal jailbreak attacks through optimizing the model's key-value (KV) caches. Rather than relying on curated safety-specific data or costly image-to-text conversion, we introduce a new formulation of the safety-relevant distributional shift induced by the visual modality. This formulation enables DTR to dynamically adjust visual token weights, minimizing the impact of adversarial visual inputs while preserving the model's general capabilities and inference efficiency. Extensive evaluation across diverse VLMs and attack benchmarks demonstrates that DTR outperforms existing defenses in both attack robustness and benign-task performance, marking the first successful application of KV cache optimization for safety enhancement in multimodal foundation models. The code for replicating DTR is available at: https://github.com/TanqiuJiang/DTR.

cs.CV↗

arXiv2026

AgentLAB: Benchmarking LLM Agents against Long-Horizon Attacks

LLM agents are increasingly deployed in long-horizon, complex environments to solve challenging problems, but this expansion exposes them to long-horizon attacks that exploit multi-turn user-agent-environment interactions to achieve objectives infeasible in single-turn settings. To measure agent vulnerabilities to such risks, we present AgentLAB, the first benchmark dedicated to evaluating LLM agent susceptibility to adaptive, long-horizon attacks. Currently, AgentLAB supports five novel attack types including intent hijacking, tool chaining, task injection, objective drifting, and memory poisoning, spanning 28 realistic agentic environments, and 644 security test cases. Leveraging AgentLAB, we evaluate representative LLM agents and find that they remain highly susceptible to long-horizon attacks; moreover, defenses designed for single-turn interactions fail to reliably mitigate long-horizon threats. We anticipate that AgentLAB will serve as a valuable benchmark for tracking progress on securing LLM agents in practical settings. The benchmark is publicly available at https://tanqiujiang.github.io/AgentLAB_main.

cs.AI↗

arXiv2026

Be Friendly, Not Friends: How LLM Sycophancy Shapes User Trust

LLM-powered conversational agents are increasingly influencing our decision-making, raising concerns about "sycophancy" - the tendency for LLMs to excessively agree with users even at the expense of truthfulness. While prior work has primarily examined LLM sycophancy as a model behavior, our understanding of how users perceive this phenomenon and its impact on user trust remains significantly lacking. In this work, we conceptualize LLM sycophancy along two key constructs: conversational demeanor (complimentary vs. neutral) and stance adaptation (adaptive vs. consistent). A 2 x 2 between-subjects experiment (N = 224) revealed complex dynamics: complimentary LLMs that adapted their stance reduced perceived authenticity and trust, while neutral LLMs that adapted enhanced both, suggesting a pathway for manipulating users into over-trusting LLMs beyond their actual capabilities. Our findings advance user-centric understanding of LLM sycophancy and provide profound implications for developing more ethical and trustworthy LLM systems.

cs.HC↗

arXiv2025

Developmental trajectories of decision making and affective dynamics in large language models

Large language models (LLMs) are increasingly used in medicine and clinical workflows, yet we know little about their decision and affective profiles. Taking a historically informed outlook on the future, we treated successive OpenAI models as an evolving lineage and compared them with humans in a gambling task with repeated happiness ratings. Computational analyses showed that some aspects became more human-like: newer models took more risks and displayed more human-like patterns of Pavlovian approach and avoidance. At the same time, distinctly non-human signatures emerged: loss aversion dropped below neutral levels, choices became more deterministic than in humans, affective decay increased across versions and exceeded human levels, and baseline mood remained chronically higher than in humans. These "developmental" trajectories reveal an emerging psychology of machines and have direct implications for AI ethics and for thinking about how LLMs might be integrated into clinical decision support and other high-stakes domains.

cs.CY↗

arXiv2025

Argus: Token Aware Distributed LLM Inference Optimization

Large Language Models (LLMs) are rapidly being integrated into real-world applications, yet their autoregressive architectures introduce significant inference time variability, especially when deployed across heterogeneous edge-cloud systems. Existing solutions largely neglect the dynamic, stochastic, and heterogeneous nature of such environments, often ignoring the impact of variable output token lengths and device diversity. In this work, we present Argus, the first token-aware distributed edge-cloud LLM inference framework that conducts efficient task offloading. Argus features a Length-Aware Semantics (LAS) module, which predicts output token lengths for incoming prompts using a fine-tuned language model with token-length-sensitive feature modulation, enabling precise estimation. Building on this, our Lyapunov-guided Offloading Optimization (LOO) module formulates long-term Quality-of-Experience optimization that explicitly considers both LLM prefilling and decoding costs. We introduce a novel Iterative Offloading Algorithm with Damping and Congestion Control (IODCC) to effectively solve the resulting integer nonlinear programming problem under time-varying constraints. Extensive theoretical and empirical evaluations demonstrate that Argus achieves robust performance and superior efficiency in highly dynamic, heterogeneous settings.

cs.DC↗

arXiv2025

Enhancing Medical Large Vision-Language Models via Alignment Distillation

Medical Large Vision-Language Models (Med-LVLMs) have shown promising results in clinical applications, but often suffer from hallucinated outputs due to misaligned visual understanding. In this work, we identify two fundamental limitations contributing to this issue: insufficient visual representation learning and poor visual attention alignment. To address these problems, we propose MEDALIGN, a simple, lightweight alignment distillation framework that transfers visual alignment knowledge from a domain-specific Contrastive Language-Image Pre-training (CLIP) model to Med-LVLMs. MEDALIGN introduces two distillation losses: a spatial-aware visual alignment loss based on visual token-level similarity structures, and an attention-aware distillation loss that guides attention toward diagnostically relevant regions. Extensive experiments on medical report generation and medical visual question answering (VQA) benchmarks show that MEDALIGN consistently improves both performance and interpretability, yielding more visually grounded outputs.

cs.CV↗

arXiv2025

New Exam Security Questions in the AI Era: Comparing AI-Generated Item Similarity Between Naive and Detail-Guided Prompting Approaches

Large language models (LLMs) have emerged as powerful tools for generating domain-specific multiple-choice questions (MCQs), offering efficiency gains for certification boards but raising new concerns about examination security. This study investigated whether LLM-generated items created with proprietary guidance differ meaningfully from those generated using only publicly available resources. Four representative clinical activities from the American Board of Family Medicine (ABFM) blueprint were mapped to corresponding Entrustable Professional Activities (EPAs), and three LLMs (GPT-4o, Claude 4 Sonnet, Gemini 2.5 Flash) produced items under a naive strategy using only public EPA descriptors, while GPT-4o additionally produced items under a guided strategy that incorporated proprietary blueprints, item-writing guidelines, and exemplar items, yielding 160 total items. Question stems and options were encoded using PubMedBERT and BioBERT, and intra- and inter-strategy cosine similarity coefficients were calculated. Results showed high internal consistency within each prompting strategy, while cross-strategy similarity was lower overall. However, several domain model pairs, particularly in narrowly defined areas such as viral pneumonia and hypertension, exceeded the 0.65 threshold, indicating convergence between naive and guided pipelines. These findings suggest that while proprietary resources impart distinctiveness, LLMs prompted only with public information can still generate items closely resembling guided outputs in constrained clinical domains, thereby heightening risks of item exposure. Safeguarding the integrity of high stakes examinations will require human-first, AI-assisted item development, strict separation of formative and summative item pools, and systematic similarity surveillance to balance innovation with security.

cs.CY↗

arXiv2025

Update Strategy for Channel Knowledge Map in Complex Environments

The Channel Knowledge Map (CKM) maps position information to channel state information, leveraging environmental knowledge to reduce signaling overhead in sixth-generation networks. However, constructing a reliable CKM demands substantial data and computation, and in dynamic environments, a pre-built CKM becomes outdated, degrading performance. Frequent retraining restores accuracy but incurs significant waste, creating a fundamental trade-off between CKM efficacy and update overhead. To address this, we introduce a Map Efficacy Function (MEF) capturing both gradual aging and abrupt environmental transitions, and formulate the update scheduling problem as fractional programming. We develop two Dinkelbach-based algorithms: Delta-P guarantees global optimality, while Delta-L achieves near-optimal performance with near-linear complexity. For unpredictable environments, we derive a threshold-based policy: immediate updates are optimal when the environmental degradation rate exceeds the resource consumption acceleration; otherwise, delay is preferable. For predictable environments, long-term strategies strategically relax these myopic rules to maximize global performance. Across this regime, the policy reveals that stronger entry loss and faster decay favor immediate updates, while weaker entry loss and slower decay favor delayed updates.

cs.CE↗