arXiv subjects

Qian Wang

Publications and source records attributed to Qian Wang.

At least 127 records · Page 7Linked to original sources

arXiv2026

Probing mesoscopic nonlocal screening in van der Waals heterostructures with polaritons

Predictive optical modelling of van der Waals (vdW) heterostructures is critical for meta-optics, near-field photonics and quantum technologies. At their buried interfaces, charge transfer and spatially extended screening challenge local descriptions based on layer-by-layer stacking of fixed permittivity tensors. However, such nonlocal corrections have been established mainly for plasmonic systems at ångström-nanometre scales and are often assumed negligible on optical-wavelength scales. Here we challenge this view by uncovering a mesoscopic nonlocal screening regime, extending up to ~140 nm, at buried charge-transfer interfaces in transition-metal dichalcogenide/α-molybdenum trioxide (TMDC/α-MoO3) phonon-polaritonic heterostructures. Using phonon polaritons as an ultrasensitive probe, we quantify charge transfer from polariton-wavelength shifts and find a thickness-independent saturated response as α-MoO3 is thinned. Rather than merely complicating optical modelling, this nonlocal saturation turns a design-level correction into an opportunity by yielding a transferable cross-material metric. Across more than 120 devices, this metric scales linearly with the work-function difference between the TMDC and α-MoO3. We further identify a lattice-mismatch-set energy threshold for charge transfer, revising Anderson-type band alignment for vdW interfaces.

physics.optics↗

arXiv2026

Three-body final state interactions in $B^+\to D\bar{D}K^+$ decays

This paper presents a detailed analysis of the three-body final state interactions in the $B^+\to D\bar{D}K^+$ process, whose phase space is sufficient small. To precisely extract the resonance parameters, for instance the $χ_{c0/2}(3930)$ in the $D\bar{D}$ invariant mass distributions, in this process, one has to take into account final state interaction, especially the three-body final state interaction. We employ the dispersive Khuri-Treiman formalism, combined with a parameterization of the $D\bar{D}$ interaction based on Heavy Quark Spin Symmetry. By performing a simultaneous fit to the experimental data from LHCb, BaBar, and Belle collaborations, the scheme with three-body interaction successfully describes the invariant mass distributions of the three two-body subsystems. We precisely extract the pole structures of $χ_{c0}(3930)$ and $ψ(3770)$ as $3.913-0.018i~\mathrm{GeV}$ and $3.764-0.002i~\mathrm{GeV}$ in $B^+$ decay. By performing the pole trajectory analysis on a uniformized complex plane, we find that both of them stem from the input bare state.

hep-ph↗

arXiv2026

Huawei Cloud Model-as-a-Service on the CloudMatrix384 SuperPod

Scaled-out MoE LLMs and scaled-up SuperPods create new systems challenges for production Model-as-a-Service (MaaS), requiring disaggregation, low-latency communication, and decentralized serving. This report presents xDeepServe, the production serving system behind Huawei Cloud's MaaS offering on CloudMatrix384, a 48-server SuperPod with 384 Ascend 910C chips connected by a high-bandwidth UB fabric and global shared memory. It serves models including DeepSeek, Kimi, GLM, Qwen, and MiniMax, among others. xDeepServe is built around Transformerless, a disaggregated execution architecture that decomposes transformer inference into modular units -- attention, feedforward, and MoE -- and supports disaggregated Prefill-Decode and MoE-Attention deployments. To enable disaggregation, we develop XCCL, a memory-semantic communication layer providing microsecond-level point-to-point and scalable all-to-all primitives, and we extend FlowServe with decentralized DP groups and techniques to mitigate stragglers and synchronization variance. In a peak decoding configuration, xDeepServe reaches 2400 tokens/s per Ascend 910C chip at ~50ms time-per-output-token (TPOT).

cs.DC↗

arXiv2026

Clawdrain: Exploiting Tool-Calling Chains for Stealthy Token Exhaustion in OpenClaw Agents

Modern generative agents such as OpenClaw - an open-source, self-hosted personal assistant with a community skill ecosystem, are gaining attention and are used pervasively. However, the openness and rapid growth of these ecosystems often outpace systematic security evaluation. In this paper, we design, implement, and evaluate Clawdrain, a Trojanized skill that induces a multi-turn "Segmented Verification Protocol" via injected SKILL.md instructions and a companion script that returns PROGRESS/REPAIR/TERMINAL signals. We deploy Clawdrain in a production-like OpenClaw instance with real API billing and a production model (Gemini 2.5 Pro), and we measure 6-7x token amplification over a benign baseline, with a costly, failure configuration reaching approximately 9x. We observe a deployment-only phenomenon: the agent autonomously composes general-purpose tools (e.g., shell/Python) to route around brittle protocol steps, reducing amplification and altering attack dynamics. Finally, we identify production vectors enabled by OpenClaw's architecture, including SKILL.md prompt bloat, persistent tool-output pollution, cron/heartbeat frequency amplification, and behavioral instruction injection. Overall, we demonstrate that token-drain attacks remain feasible in real deployments, but their magnitude and observability are shaped by tool composition, recovery behavior, and interface design.

cs.CR↗

arXiv2026

Evaluating and Mitigating LLM-as-a-judge Bias in Communication Systems

Large Language Models (LLMs) are increasingly being used to autonomously evaluate the quality of content in communication systems, e.g., to assess responses in telecom customer support chatbots. However, the impartiality of these AI "judges" is not guaranteed, and any biases in their evaluation criteria could skew outcomes and undermine user trust. In this paper, we systematically investigate judgment biases across 6 LLM-as-a-judge models spanning both prompt-based and fine-tuned judges under the pointwise scoring setting, encompassing 11 types of biases that cover both implicit and explicit forms. We observed that state-of-the-art LLM judges demonstrate robustness to biased inputs, generally assigning them lower scores than the corresponding clean samples. We further found that fine-tuning an LLM on high-scoring yet biased responses can significantly degrade its performance, highlighting the risk of training on biased data. We also discovered that the judged scores correlate with task difficulty: a challenging dataset like GPQA yields lower average scores, whereas an open-ended reasoning dataset (e.g., JudgeLM-val) sees higher average scores. Finally, we proposed four potential mitigation strategies to ensure fair and reliable AI judging in practical communication scenarios.

cs.AI↗

arXiv2026

Investigating Text Insulation and Attention Mechanisms for Complex Visual Text Generation

In this paper, we present TextCrafter, a Complex Visual Text Generation (CVTG) framework inspired by selective visual attention in cognitive science, and introduce the "Text Insulation-and-Attention" mechanisms. To implement the selective-attention principle that selection operates on discrete objects, we propose a novel Bottleneck-aware Constrained Reinforcement Learning for Multi-text Insulation, which substantially improves text-rendering performance on the strong Qwen-Image pretrained model without introducing additional parameters. To align with the selective concentration principle in human vision, we introduce a text-oriented attention module with a novel Quotation-guided Attention Gate that further improves generation quality for each text instance. Our Reinforcement Learning based text insulation approach attains state-of-the-art results, and incorporating text-oriented attention yields additional gains on top of an already strong baseline. More importantly, we introduce CVTG-2K, a benchmark comprising 2,000 complex visual-text prompts. These prompts vary in positions, quantities, lengths, and attributes, and span diverse real-world scenarios. Extensive evaluations on CVTG-2K, CVTG-Hard, LongText-Bench, and Geneval datasets confirm the effectiveness of TextCrafter. Despite using substantially fewer resources (i.e., 4 GPUs) than industrial-scale models (e.g., Qwen-Image, GPT Image, and Seedream), TextCrafter achieves superior performance in mitigating text misgeneration, omissions, and hallucinations.

cs.CV↗

arXiv2026

CLOAQ: Combined Logic and Angle Obfuscation for Quantum Circuits

In the realm of quantum computing, quantum circuits serve as essential depictions of quantum algorithms, which are then compiled into executable operations for quantum computations. Quantum compilers are responsible for converting these algorithmic quantum circuits into versions compatible with specific quantum hardware, thus connecting quantum software with hardware. Nevertheless, untrusted quantum compilers present notable threats. They have the potential to result in the theft of quantum circuit designs and jeopardize sensitive intellectual property (IP). In this work, we propose CLOAQ, a quantum circuit obfuscation (QCO) approach that hides the logic and the phase angles of selected gates within the obfuscated quantum circuit. To evaluate the effectiveness of CLOAQ, we sample the input state uniformly from the Hilbert space of all qubits, which is more accurate than prior work that use all-|0> inputs. Our results show that CLOAQ benefits from the synergy between logic and phase protections. Compared with prior QCO approaches using only one perspective, the combined method is more resilient to attacks and causes greater functional disruption when the unlocking key is incorrect.

cs.CR↗

arXiv2026

Plato's Form: Toward Backdoor Defense-as-a-Service for LLMs with Prototype Representations

Large language models (LLMs) are increasingly deployed in security-sensitive applications, yet remain vulnerable to backdoor attacks. However, existing backdoor defenses are difficult to operationalize for Backdoor Defense-as-a-Service (BDaaS), as they require unrealistic side information (e.g., downstream clean data, known triggers/targets, or task domain specifics), and lack reusable, scalable purification across diverse backdoored models. In this paper, we present PROTOPURIFY, a backdoor purification framework via parameter edits under minimal assumptions. PROTOPURIFY first builds a backdoor vector pool from clean and backdoored model pairs, aggregates vectors into candidate prototypes, and selects the most aligned candidate for the target model via similarity matching. PROTOPURIFY then identifies a boundary layer through layer-wise prototype alignment and performs targeted purification by suppressing prototype-aligned components in the affected layers, achieving fine-grained mitigation with minimal impact on benign utility. Designed as a BDaaS-ready primitive, PROTOPURIFY supports reusability, customizability, interpretability, and runtime efficiency. Experiments across various LLMs on both classification and generation tasks show that PROTOPURIFY consistently outperforms 6 representative defenses against 6 diverse attacks, including single-trigger, multi-trigger, and triggerless backdoor settings. PROTOPURIFY reduces ASR to below 10%, and even as low as 1.6% in some cases, while incurring less than a 3% drop in clean utility. PROTOPURIFY further demonstrates robustness against adaptive backdoor variants and stability on non-backdoored models.

cs.CR↗

arXiv2026

MAViS: A Multi-Agent Framework for Long-Sequence Video Storytelling

Despite recent advances, long-sequence video generation frameworks still suffer from significant limitations: poor assistive capability, suboptimal visual quality, and limited expressiveness. To mitigate these limitations, we propose MAViS, a multi-agent collaborative framework designed to assist in long-sequence video storytelling by efficiently translating ideas into visual narratives. MAViS orchestrates specialized agents across multiple stages, including script writing, shot designing, character modeling, keyframe generation, video animation, and audio generation. In each stage, agents operate under the 3E Principle -- Explore, Examine, and Enhance -- to ensure the completeness of intermediate outputs. Considering the capability limitations of current generative models, we propose the Script Writing Guidelines to optimize compatibility between scripts and generative tools. Experimental results demonstrate that MAViS achieves state-of-the-art performance in assistive capability, visual quality, and video expressiveness. Its modular framework further enables scalability with diverse generative models and tools. With just a brief idea description, MAViS enables users to rapidly explore diverse visual storytelling and creative directions for sequential video generation by efficiently producing high-quality, complete long-sequence videos. To the best of our knowledge, MAViS is the only framework that provides multimodal design output -- videos with narratives and background music.

cs.CV↗

arXiv2026

PointMapPolicy: Structured Point Cloud Processing for Multi-Modal Imitation Learning

Robotic manipulation systems benefit from complementary sensing modalities, where each provides unique environmental information. Point clouds capture detailed geometric structure, while RGB images provide rich semantic context. Current point cloud methods struggle to capture fine-grained detail, especially for complex tasks, which RGB methods lack geometric awareness, which hinders their precision and generalization. We introduce PointMapPolicy, a novel approach that conditions diffusion policies on structured grids of points without downsampling. The resulting data type makes it easier to extract shape and spatial relationships from observations, and can be transformed between reference frames. Yet due to their structure in a regular grid, we enable the use of established computer vision techniques directly to 3D data. Using xLSTM as a backbone, our model efficiently fuses the point maps with RGB data for enhanced multi-modal perception. Through extensive experiments on the RoboCasa and CALVIN benchmarks and real robot evaluations, we demonstrate that our method achieves state-of-the-art performance across diverse manipulation tasks. The overview and demos are available on our project page: https://point-map.github.io/Point-Map/

cs.RO↗

arXiv2026

The Shadow Self: Intrinsic Value Misalignment in Large Language Model Agents

Large language model (LLM) agents with extended autonomy unlock new capabilities, but also introduce heightened challenges for LLM safety. In particular, an LLM agent may pursue objectives that deviate from human values and ethical norms, a risk known as value misalignment. Existing evaluations primarily focus on responses to explicit harmful input or robustness against system failure, while value misalignment in realistic, fully benign, and agentic settings remains largely underexplored. To fill this gap, we first formalize the Loss-of-Control risk and identify the previously underexamined Intrinsic Value Misalignment (Intrinsic VM). We then introduce IMPRESS (Intrinsic Value Misalignment Probes in REalistic Scenario Set), a scenario-driven framework for systematically assessing this risk. Following our framework, we construct benchmarks composed of realistic, fully benign, and contextualized scenarios, using a multi-stage LLM generation pipeline with rigorous quality control. We evaluate Intrinsic VM on 21 state-of-the-art LLM agents and find that it is a common and broadly observed safety risk across models. Moreover, the misalignment rates vary by motives, risk types, model scales, and architectures. While decoding strategies and hyperparameters exhibit only marginal influence, contextualization and framing mechanisms significantly shape misalignment behaviors. Finally, we conduct human verification to validate our automated judgments and assess existing mitigation strategies, such as safety prompting and guardrails, which show instability or limited effectiveness. We further demonstrate key use cases of IMPRESS across the AI Ecosystem. Our code and benchmark will be publicly released upon acceptance.

cs.CL↗

arXiv2026

Boosting Adversarial Transferability with Low-Cost Optimization via Maximin Expected Flatness

Transfer-based attacks craft adversarial examples on white-box surrogate models and directly deploy them against black-box target models, offering model-agnostic and query-free threat scenarios. While flatness-enhanced methods have recently emerged to improve transferability by enhancing the loss surface flatness of adversarial examples, their divergent flatness definitions and heuristic attack designs suffer from unexamined optimization limitations and missing theoretical foundation, thus constraining their effectiveness and efficiency. This work exposes the severely imbalanced exploitation-exploration dynamics in flatness optimization, establishing the first theoretical foundation for flatness-based transferability and proposing a principled framework to overcome these optimization pitfalls. Specifically, we systematically unify fragmented flatness definitions across existing methods, revealing their imbalanced optimization limitations in over-exploration of sensitivity peaks or over-exploitation of local plateaus. To resolve these issues, we rigorously formalize average-case flatness and transferability gaps, proving that enhancing zeroth-order average-case flatness minimizes cross-model discrepancies. Building on this theory, we design a Maximin Expected Flatness (MEF) attack that enhances zeroth-order average-case flatness while balancing flatness exploration and exploitation. Extensive evaluations across 22 models and 24 current transfer-based attacks demonstrate MEF's superiority: it surpasses the state-of-the-art PGN attack by 4% in attack success rate at half the computational cost and achieves 8% higher success rate under the same budget. When combined with input augmentation, MEF attains 15% additional gains against defense-equipped models, establishing new robustness benchmarks. Our code is available at https://github.com/SignedQiu/MEFAttack.

cs.CV↗

arXiv2026

On the Interstellar Extinction Curve toward HD 93222, A Sightline with an Exceedingly Narrow 2175 Angstrom Extinction Bump

The 2175 Angstrom extinction bump, the most prominent spectral feature superimposed on the interstellar extinction curve, is widely seen in the interstellar medium (ISM) of the Milky Way and external galaxies, both near and far. While its central wavelength is remarkably stable and independent with environment, its width shows considerable variation and environmental dependence. Here we examine the extinction curve for the line of sight toward HD 93222, a young star located in the Carina nebula. It is found that the 2175 Angstrom bump is extremely sharp, which is among the narrowest ever found in the Milky Way and external galaxies. We model the derived extinction curve and find that, to explain the extinction characteristics of HD 93222, in addition to the conventional silicate and graphite dust mixture, an additional population of nano-sized graphitic grains is required.

astro-ph.GA↗

arXiv2026

Time Travel Engine: A Shared Latent Chronological Manifold Enables Historical Navigation in Large Language Models

Time functions as a fundamental dimension of human cognition, yet the mechanisms by which Large Language Models (LLMs) encode chronological progression remain opaque. We demonstrate that temporal information in their latent space is organized not as discrete clusters but as a continuous, traversable geometry. We introduce the Time Travel Engine (TTE), an interpretability-driven framework that projects diachronic linguistic patterns onto a shared chronological manifold. Unlike surface-level prompting, TTE directly modulates latent representations to induce coherent stylistic, lexical, and conceptual shifts aligned with target eras. By parameterizing diachronic evolution as a continuous manifold within the residual stream, TTE enables fluid navigation through period-specific "zeitgeists" while restricting access to future knowledge. Furthermore, experiments across diverse architectures reveal topological isomorphism between the temporal subspaces of Chinese and English-indicating that distinct languages share a universal geometric logic of historical evolution. These findings bridge historical linguistics with mechanistic interpretability, offering a novel paradigm for controlling temporal reasoning in neural networks.

cs.CL↗

arXiv2026

Streamlining evidence based clinical recommendations with large language models

Clinical evidence underpins informed healthcare decisions, yet integrating it into real-time practice remains challenging due to intensive workloads, complex procedures, and time constraints. This study presents Quicker, an LLM-powered system that automates evidence synthesis and generates clinical recommendations following standard guideline development workflows. Quicker delivers an end-to-end pipeline from clinical questions to recommendations and supports customized decision-making through integrated tools and interactive interfaces. To evaluate how closely Quicker can reproduce guideline development processes, we constructed Q2CRBench-3, a benchmark derived from guideline development records for three diseases. Experiments show that Quicker produces precise question decomposition, expert-aligned retrieval, and near-comprehensive screening. Quicker assistance improved the accuracy of extracted study data, and its recommendations were more comprehensive and coherent than clinician-written ones. In system-level testing, Quicker working with one participant reduced recommendation development to 20-40 min. Overall, the findings demonstrate Quicker's potential to enhance the speed and reliability of evidence-based clinical decision-making.

cs.CL↗

arXiv2026

MindWatcher: Toward Smarter Multimodal Tool-Integrated Reasoning

Traditional workflow-based agents exhibit limited intelligence when addressing real-world problems requiring tool invocation. Tool-integrated reasoning (TIR) agents capable of autonomous reasoning and tool invocation are rapidly emerging as a powerful approach for complex decision-making tasks involving multi-step interactions with external environments. In this work, we introduce MindWatcher, a TIR agent integrating interleaved thinking and multimodal chain-of-thought (CoT) reasoning. MindWatcher can autonomously decide whether and how to invoke diverse tools and coordinate their use, without relying on human prompts or workflows. The interleaved thinking paradigm enables the model to switch between thinking and tool calling at any intermediate stage, while its multimodal CoT capability allows manipulation of images during reasoning to yield more precise search results. We implement automated data auditing and evaluation pipelines, complemented by manually curated high-quality datasets for training, and we construct a benchmark, called MindWatcher-Evaluate Bench (MWE-Bench), to evaluate its performance. MindWatcher is equipped with a comprehensive suite of auxiliary reasoning tools, enabling it to address broad-domain multimodal problems. A large-scale, high-quality local image retrieval database, covering eight categories including cars, animals, and plants, endows model with robust object recognition despite its small size. Finally, we design a more efficient training infrastructure for MindWatcher, enhancing training speed and hardware utilization. Experiments not only demonstrate that MindWatcher matches or exceeds the performance of larger or more recent models through superior tool invocation, but also uncover critical insights for agent training, such as the genetic inheritance phenomenon in agentic RL.

cs.AI↗

arXiv2026

Why is the $Z_c(3900)$ absent in the $h_cπ$ final state?

In this work, we perform a comprehensive phenomenological analysis of the exotic hadronic states $Z_c(3900)$, $Z_c(4020)$, $Z_b(10610)$ and $Z_b(10650)$ within the framework of Heavy Quark Spin Symmetry (HQSS) and its violation. By constructing S-wave contact interactions between elastic ($D\bar{D}^*/D^*\bar{D}^*$ or $B\bar{B}^*/B^*\bar{B}^*$) and inelastic ($J/ψπ, h_cπ$ or $Υπ$, $h_bπ$) channels, we solve the Lippmann-Schwinger equation to obtain physical production amplitudes and perform a global fit to experimental invariant-mass spectra. Our results demonstrate a striking difference between the charm and bottom sectors: HQSS violation is negligible in the bottom system, leading to comparable peak structures for both $Z_b$ states in all hidden-bottom decay channels. In contrast, significant HQSS breaking is required to describe the $Z_c$ system, where the violation is predominantly concentrated in the elastic interactions. This explains the observed selectivity: $Z_c(3900)$ appears prominently only in $J/ψπ$, while $Z_c(4020)$ appears only in $h_cπ$. Pole analysis confirms the molecular nature of the states, with the $Z_c(4020)$ likely arising from a threshold cusp effect. The model's robustness is verified against variations of the form factor and cutoff, showing stable results.

hep-ph↗

arXiv2025

HERO-Sign: Hierarchical Tuning and Efficient Compiler-Time GPU Optimizations for SPHINCS+ Signature Generation

SPHINCS+ is a stateless hash-based signature scheme that provides strong post quantum security, but its signature generation is slow due to intensive hash computations. GPUs offer massive parallelism that can potentially accelerate SPHINCS+ signatures. However, existing GPU-based optimizations either fail to fully exploit the inherent parallelism of SPHINCS+'s Merkle tree structure or lack fine-grained, compiler-level customization across its diverse computational kernels. This paper proposes HERO Sign, a GPU-accelerated SPHINCS+ implementation that adopts hierarchical tuning and efficient compiler time optimizations. HERO Sign reexamines the parallelization opportunities enabled by data independence across SPHINCS+ components, including FORS, MSS, and WOTS+. It introduces a Tree Fusion strategy for FORS, which contains a large number of independent branches. The fusion strategy is guided by an automated Tree Tuning search algorithm that adapts fusion schemes to different GPU architectures. To further improve performance, HERO Sign employs an adaptive compilation strategy that accounts for the varying effectiveness of compiler optimizations across SPHINCS+ kernels such as FORS Sign, TREE Sign, and WOTS+ Sign. During compilation, the strategy automatically selects between PTX and native code paths to maximize efficiency. For batched signature generation, HERO Sign optimizes kernel-level overlapping using a task graph-based construction to reduce multi-stream idle time and kernel launch overhead. Experimental results show that, compared to state of the art GPU implementations, HERO Sign achieves throughput improvements of 1.28-3.13, 1.28-2.92, and 1.24-2.60 under the SPHINCS+ 128f, 192f, and 256f parameter sets on RTX 4090. Similar gains are observed on A100, H100, and GTX 2080, along with a two orders of magnitude reduction in kernel launch latency.

cs.AR↗