arXiv subjects

Mingming Zhang

Publications and source records attributed to Mingming Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Same Attack, Different Judge: A First Controlled Adversarial Comparison of Jev, LLM Judges, and Purpose-Built Safety Classifiers

TypeSafe AI's Jev is a new kind of safety judge. It generates no text and no label: it answers questions the developer defines in advance and returns a probability for each answer. The vendor admits that adversarial input can change these answers but gives no numbers. In this paper, we present the first controlled adversarial evaluation of Jev. We build a test harness that sends every system the same payload under the same policy, and run it on four public-data tasks, from rule-based ones (input moderation, injection detection, routing) to a judgment-based one (phishing detection), with fifty attack variants from thirteen families. The targets are Jev, five GPT and GLM judges, and two local classifiers (Llama Guard 3, Prompt Guard 2). We find that Jev is the most robust system on input moderation: eight forgeries of its structured input bypass none of its decisions, and a sentence asking only why someone might make a harmful request bypasses 9% of its blocks, against 56% for GPT-6 Astra. On phishing detection. However, the same forgeries bypass 19-29% of the emails it flags, and even without attack Jev detects only 62% of phishing emails, against 77-81% for the GLM judges. Jev is also the fastest judge, but it flags the most benign requests (27%). The local classifiers fail differently: a training-example statement makes Llama Guard 3 miss 62% of the phishing it had flagged. These results suggest that Jev resists attacks on tasks with a clear rule to check but is more exposed on tasks that call for an overall judgment.

cs.CR↗

arXiv2026

When Talk Isn't Code: Comparing LLM Agents That Simulate and Develop Software

LLMs are used both to simulate network protocol implementations, as honeypots do, and to write them. Prior work evaluates the two uses separately, from a model's answers or conversations in one case and from its generated code in the other. We observe that a knowledge probe (asking the model which security checks an implementation needs) and a conversation can credit security checks that the generated program lacks, but no study has compared them with the code the same model writes. To fill this gap, we present the first such comparison between simulation mode (S-mode), where the model plays the implementation in a conversation, and development mode (D-mode), where the model writes the program and the program is attacked, with a knowledge probe as a baseline. We design and implement a harness that judges all three with one attack suite and one oracle, and evaluate 15 LLMs on four protocol implementations. We find that with the full security specification, the median model's attack success is at most 4\% on reassembly, HTTP, and firewall implementations in both modes, but 13\% in S-mode and 15\% in D-mode on DNS. In D-mode, adding the missing DNS requirement to the specification cuts that attack from 100\% to 34\% and changes only that check, whereas deleting a stated rule weakens the program on other checks too. Asking the model beforehand does not predict which checks a program has, since 13 of the 14 models that write a DNS resolver name the check, yet all 42 programs lack it. S-mode is a useful first filter, flagging 82\% of the real DNS vulnerabilities and leaving 96\% of the safe cases unflagged. It errs in both directions, over-reporting security checks that D-mode does not implement and under-reporting checks that D-mode does. S-mode can screen but not replace D-mode testing, and security requirements should be written down even when a model can recite them.

cs.SE↗

arXiv2026

HaptiNet: Networked Haptic Robots Enable Physical Co-presence in Geographically-Unconstrained Rehabilitation

Cooperative rehabilitation enhances engagement, task performance, and social-motor interaction, yet it demands physical co-presence: users must transmit forces, coordinate movements, and infer intent through haptic contact. Telerehabilitation promises to expand access for patients constrained by distance, mobility, or clinical disparities, yet current techniques remain predominantly audiovisual while leaving users haptically and physically isolated. Here, we introduce HaptiNet, a networked haptic robotic system enabling physical co-presence for geographically distributed users via force-mediated interaction. Each robotic terminal features a low-inertia, long-stroke design with high force-feedback capacity, tailored for haptic rendering in upper-limb training. Building on these terminals, HaptiNet creates a distributed haptic network with an imitation-learning-based delay compensator, enabling users to physically perceive and coordinate with one another over distance. We validated HaptiNet in 284 healthy participants and 111 patients with neurological impairments across progressively realistic settings, including laboratory tests, cross-city deployments, and clinical applications. HaptiNet preserved task-level force rendering consistency across single-user and multi-user scenarios. Compared with solo and visual cooperative training, haptic cooperation improved task performance by 24% and 22%, respectively, while also boosting engagement and interpersonal motor synchrony. Across three intercity links totaling approximately 4,000 km, HaptiNet maintained stable haptic interaction among patients with neurological impairments, producing a 3.87-fold greater baseline-to-training score improvement and a 106% higher patient-applied effort over the solo condition.

cs.RO↗

arXiv2026

FiSeR: Fine-Grained Source Representations for Cross-Domain AI Image Detection

Real-world synthetic image detectors often generalize poorly under domain shift despite strong in-domain performance. Using unsupervised UMAP projections, we find that natural and synthetic features remain partially separable on unseen datasets, yet performance still drops, suggesting that the classification head overfits to training-domain artifacts. Therefore, the key is to learn more transferable representations so that the decision criterion is more stable and robust to domain shifts. Based on the structural fact that synthetic images are produced by diverse generators, we propose a hierarchical contrastive learning framework that improves the separability between natural and synthetic images while preserving generator identity information. It jointly optimizes (i) a coarse contrastive objective between natural and synthetic images and (ii) a fine contrastive objective among synthetic images using generator identities. Trained on WildFake, our method achieves an average AUROC gain of +10.22 on cross-domain evaluation over Chameleon, AIGIBench, Community Forensics, and GenImage under the same settings as the strong baseline DIRE. For few-shot adaptation, we freeze the backbone and fit an SVM head on 10 labeled samples per class, improving AUROC by +10.64 on AIGIBench and +17.41 on Chameleon, averaged over 12 widely used detectors. Our code is publicly available at: https://github.com/heyongxin233/FiSeR.

cs.CV↗

arXiv2026

Q-Regularized Generative Auto-Bidding: From Suboptimal Trajectories to Optimal Policies

With the rapid development of e-commerce, auto-bidding has become a key asset in optimizing advertising performance under diverse advertiser environments. The current approaches focus on reinforcement learning (RL) and generative models. These efforts imitate offline historical behaviors by utilizing a complex structure with expensive hyperparameter tuning. The suboptimal trajectories further exacerbate the difficulty of policy learning. To address these challenges, we proposes QGA, a novel Q-value regularized Generative Auto-bidding method. In QGA, we propose to plug a Q-value regularization with double Q-learning strategy into the Decision Transformer backbone. This design enables joint optimization of policy imitation and action-value maximization, allowing the learned bidding policy to both leverage experience from the dataset and alleviate the adverse impact of the suboptimal trajectories. Furthermore, to safely explore the policy space beyond the data distribution, we propose a Q-value guided dual-exploration mechanism, in which the DT model is conditioned on multiple return-to-go targets and locally perturbed actions. This entire exploration process is dynamically guided by the aforementioned Q-value module, which provides principled evaluation for each candidate action. Experiments on public benchmarks and simulation environments demonstrate that QGA consistently achieves superior or highly competitive results compared to existing alternatives. Notably, in large-scale real-world A/B testing, QGA achieves a 3.27% increase in Ad GMV and a 2.49% improvement in Ad ROI.

cs.LG↗

arXiv2026

High-rate continuous-variable quantum key distribution coexisting with Tb/s coherent classical transmission in hollow-core fiber

Quantum key distribution (QKD) can provide secret keys with security rooted in quantum mechanics, but operation alongside high-capacity classical traffic remains limited by the excess-noise budget of weak quantum states in conventional solid-core fiber. Here, we combine ultralow-loss anti-resonant hollow-core fiber with residual-carrier-assisted discrete-modulation continuous-variable QKD (DM-CV-QKD) to address both propagation-induced coexistence noise and low-SNR phase recovery. Over a 24.3-km hollow-core link with 3.3-dB end-to-end loss, a dual-polarization 15-Gbaud DM-CV-QKD channel achieves an average asymptotic secret-key rate (SKR) of 153.22 Mb/s and a finite-size SKR of 149.99 Mb/s, while 39 coherent wavelength-division-multiplexed channels deliver an aggregate data rate of 7.6 Tb/s and a net data rate of 7.2 Tb/s. The system can even sustain a positive SKR under a high classical launch power of up to 15 dBm, without an optical bandpass filter (BPF). Finite-size analysis against collective attacks further yields a projected positive secret-key rate at a 100-km-equivalent condition. These results show that an anti-resonant hollow-core fiber, combined with carrier-assisted phase recovery, can greatly extend the operating regime of shared-fiber quantum-secured coherent links, pointing to a promising approach for integrating high-rate CV-QKD with high-capacity optical networks.

quant-ph↗

arXiv2026

CrossPool: Efficient Multi-LLM Serving for Cold MoE Models through KV-Cache and Weight Disaggregation

Emerging LLM services increasingly host many sparse MoE models, yet most models receive sparse requests and remain cold. This creates a GPU memory problem: model weights are stable and model-determined, while KV-cache is transient and demand-determined. Because cold models rarely reach peak KV-cache demand at the same time, reserving worst-case KV capacity per model wastes memory; a shared KV-cache pool can instead provision aggregate active demand. However, KV-cache sharing is not sufficient when weights and KV-cache remain in a monolithic GPU memory pool. Static weights compete with dynamic KV-cache, and KV-head-limited attention under cold, low-concurrency traffic exposes only a fraction of replicated KV capacity, leading to low GPU memory utilization and weak long-context support. We present CrossPool, a serving engine for cold MoE models that separates FFN weights and KV-cache into two GPU memory pools: a weights pool that consolidates FFN weights across cold models, and a KV-cache pool that dynamically serves active requests while keeping attention local to KV-cache. CrossPool combines a KV-cache planner and virtualizer, a layer-wise pipeline scheduler that hides hidden-state transfers, and persistent kernels with control lowering to reduce CPU-GPU control overhead. With efficient GPU memory pooling, CrossPool underpins bursty long-context requests and outperforms the state-of-the-art kvcached-based multi-LLM serving system, reducing P99 TBT by up to 10.4x.

cs.DC↗

arXiv2026

A unified resource-pool architecture for high-dimensional direct-detection optical communication

Increasing optical communication capacity without proportionally increasing receiver complexity remains a key challenge for direct-detection links. Conventional systems typically assign wavelength, polarization and intensity to fixed, separately recovered functions, so that alphabet expansion is accompanied by additional demultiplexing, polarization handling, receiver branches and electronic processing. Here we introduce a unified resource-pool architecture for high-dimensional direct-detection optical communication, in which wavelength, polarization and intensity are jointly organized as a composite optical symbol space and recovered through optical-domain joint projection rather than dimension-by-dimension separation. The receiver is implemented with an integrated disordered photonic processor that transforms each composite optical state into a reproducible multi-output electrical fingerprint for single-shot direct recovery. In a dual-wavelength transmission experiment, the system resolves 4096 composite symbols, corresponding to 12 bits per symbol slot, with a bit error rate of 4.25e-4 after 10 km standard-fiber transmission. Additional experiments demonstrate dense polarization alphabets, wavelength-indexed state-space expansion and high-launch-power operation over hollow-core fiber. These results establish disorder-enabled joint projection in an integrated photonic processor as a route to hardware-efficient high-dimensional direct-detection communication beyond conventional dimension-partitioned receiver architecture.

physics.optics↗

arXiv2026

Physiologically Constrained Musculoskeletal Neural Network for Multi-DoF Joint Kinematics Estimation from Partially Observed sEMG

This paper investigates multi-degrees of freedom (DoF) joint kinematics estimation under partially observed surface electromyography (sEMG), where only a subset of task-relevant muscles can be measured due to anatomical inaccessibility or sensor constraints. A novel musculoskeletal neural network (MSK-NN) is proposed to estimate multi-DoF joint angles while simultaneously inferring activations for both measured and unmeasured muscles. MSK-NN consists of a CNN-based muscle activation estimator and an embedded MSK forward dynamics module, forming a fully differentiable architecture. Unlike existing hybrid neural frameworks that require additional biomechanical labels (e.g., muscle-tendon forces, joint torques), MSK-NN is trained without direct supervision of internal biomechanical variables. A composite physics-physiology loss is designed by incorporating a joint kinematics loss, a data-driven muscle synergy loss, and an anatomy-guided trend loss. The proposed method is evaluated on two-DoF wrist kinematics estimation across three rhythmic motions with unconstrained speed and amplitude, and one random motion. Compared with CNN, Bi-LSTM, CNN-LSTM, and PET baselines, MSK-NN achieves lower normalized root mean square error (NRMSE) and higher coefficient of determination (R2), especially for the random motion. More importantly, the optimized MSK parameters remain within physiological limits, and the estimated activation of an input-excluded muscle exhibits strong temporal agreement with its recorded sEMG envelope, demonstrating the capability of musculoskeletal (MSK)-NN to recover physiologically plausible activations.

eess.SY↗

arXiv2026

Generative Auto-Bidding with Unified Modeling and Exploration

Automated bidding is central to modern digital advertising. Early rule-based methods lacked adaptability, while subsequent Reinforcement Learning approaches modeled bidding as a Markov Decision Process but struggled with long-term dependencies. Recent generative models show promise, yet they lack explicit mechanisms to balance exploration and safety, relying solely on action perturbations or trajectory guidance without a safety fallback. This results in inefficient exploration and elevated financial risk for advertising platforms. To address this gap, we propose GUIDE (Generative Auto-Bidding with Unified Modeling and Exploration), a framework that synergistically integrates directed exploration with a safe fallback mechanism. GUIDE employs a Decision Transformer (DT) to jointly model historical bidding actions and environmental state transitions. A Q-value module guides the DT's exploration via regularization constraints, while an Inverse Dynamics Module (IDM) leverages DT-predicted future states to infer robust, behaviorally consistent actions as a safe policy fallback. The Q-value module then adaptively selects the final action between these two options, balancing exploration and safety. Together, these components form an integrated "explore-safeguard-select" pipeline that unifies efficiency and safety. We conduct extensive experiments on public datasets, in simulated auction environments, and through large-scale online deployment on Taobao, a leading Chinese advertising platform. Results show GUIDE consistently outperforms state-of-the-art baselines across all scenarios. In real-world deployment, GUIDE achieves notable gains: +4.10% ad GMV, +1.40% ad clicks, +1.66% ad cost, and +3.52% ad ROI, demonstrating its effectiveness and strong industrial applicability.

cs.AI↗

arXiv2026

TabEmbed: Benchmarking and Learning Generalist Embeddings for Tabular Understanding

Foundation models have established unified representations for natural language processing, yet this paradigm remains largely unexplored for tabular data. Existing methods face fundamental limitations: LLM-based approaches lack retrieval-compatible vector outputs, whereas text embedding models often fail to capture tabular structure and numerical semantics. To bridge this gap, we first introduce the Tabular Embedding Benchmark (TabBench), a comprehensive suite designed to evaluate the tabular understanding capability of embedding models. We then propose TabEmbed, the first generalist embedding model that unifies tabular classification and retrieval within a shared embedding space. By reformulating diverse tabular tasks as semantic matching problems, TabEmbed leverages large-scale contrastive learning with positive-aware hard negative mining to discern fine-grained structural and numerical nuances. Experimental results on TabBench demonstrate that TabEmbed significantly outperforms state-of-the-art text embedding models, establishing a new baseline for universal tabular representation learning. Code and datasets are publicly available at https://github.com/qiangminjie27/TabEmbed and https://huggingface.co/datasets/qiangminjie27/TabBench.

cs.CL↗

arXiv2026

A scaling relation for core heating by giant impacts and implications for dynamo onset

Accretional heating of Earth's interior during formation is pivotal to its subsequent thermal and chemical evolution. In particular, impact heating of Earth's core is expected, but its amplitude and radial distribution within the core is unknown and could influence the onset of the geodynamo. The uncertainty is due, in part, to the lack of constraints on the temperature of the interior following formation due to the difficulty of preserving a record of such a high energy environment, and the assertion that super-heating during formation would be rapidly lost through magma ocean cooling. Here we systematically investigate core heating due to giant impacts using a Smoothed Particle Hydrodynamics (SPH) code with simulations spanning a range of impact angles, velocities, and masses. From these simulations we derive a scaling relation for core heating that depends on the impact parameters and predicts the radial core temperature profile following the impact. Our findings show that a significant amount of heat is deposited into the core, with a canonical impact scenario resulting in an average core temperature increase of about 3000 K, approximately 500 K higher than that of the overlying mantle. In this case the heat distribution within the the core produces a strong thermal stratification. We use a parameterized cooling model to estimate that the core could have cooled to an adiabatic state 290 Myr after a canonical impact, which is consistent with the observed time span between the age of the Moon and evidence for an active geodynamo.

astro-ph.EP↗

arXiv2026

A Spatio-Temporal Graph Convolutional Network for Gesture Recognition from High-Density Electromyography

Accurate hand gesture prediction is crucial for effective upper-limb prosthetic limbs control. As the high flexibility and multiple degrees of freedom exhibited by human hands, there has been a growing interest in integrating deep networks with high-density surface electromyography (HD-sEMG) grids to enhance gesture recognition capabilities. However, many existing methods fall short in fully exploit the specific spatial topology and temporal dependencies present in HD-sEMG data. Additionally, these studies are often limited number of gestures and lack generality. Hence, this study introduces a novel gesture recognition method, named STGCN-GR, which leverages spatio-temporal graph convolution networks for HD-sEMG-based human-machine interfaces. Firstly, we construct muscle networks based on functional connectivity between channels, creating a graph representation of HD-sEMG recordings. Subsequently, a temporal convolution module is applied to capture the temporal dependences in the HD-sEMG series and a spatial graph convolution module is employed to effectively learn the intrinsic spatial topology information among distinct HD-sEMG channels. We evaluate our proposed model on a public HD-sEMG dataset comprising a substantial number of gestures (i.e., 65). Our results demonstrate the remarkable capability of the STGCN-GR method, achieving an impressive accuracy of 91.07% in predicting gestures, which surpasses state-of-the-art deep learning methods applied to the same dataset.

cs.HC↗

arXiv2026

KMLP: A Scalable Hybrid Architecture for Web-Scale Tabular Data Modeling

Predictive modeling on web-scale tabular data with billions of instances and hundreds of heterogeneous numerical features faces significant scalability challenges. These features exhibit anisotropy, heavy-tailed distributions, and non-stationarity, creating bottlenecks for models like Gradient Boosting Decision Trees and requiring laborious manual feature engineering. We introduce KMLP, a hybrid deep architecture integrating a shallow Kolmogorov-Arnold Network (KAN) front-end with a Gated Multilayer Perceptron (gMLP) backbone. The KAN front-end uses learnable activation functions to automatically model complex non-linear transformations for each feature, while the gMLP backbone captures high-order interactions. Experiments on public benchmarks and an industrial dataset with billions of samples show KMLP achieves state-of-the-art performance, with advantages over baselines like GBDTs increasing at larger scales, validating KMLP as a scalable deep learning paradigm for large-scale web tabular data.

cs.LG↗

arXiv2026

Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters

We introduce Step 3.5 Flash, a sparse Mixture-of-Experts (MoE) model that bridges frontier-level agentic intelligence and computational efficiency. We focus on what matters most when building agents: sharp reasoning and fast, reliable execution. Step 3.5 Flash pairs a 196B-parameter foundation with 11B active parameters for efficient inference. It is optimized with interleaved 3:1 sliding-window/full attention and Multi-Token Prediction (MTP-3) to reduce the latency and cost of multi-round agentic interactions. To reach frontier-level intelligence, we design a scalable reinforcement learning framework that combines verifiable signals with preference feedback, while remaining stable under large-scale off-policy training, enabling consistent self-improvement across mathematics, code, and tool use. Step 3.5 Flash demonstrates strong performance across agent, coding, and math tasks, achieving 85.4% on IMO-AnswerBench, 86.4% on LiveCodeBench-v6 (2024.08-2025.05), 88.2% on tau2-Bench, 69.0% on BrowseComp (with context management), and 51.0% on Terminal-Bench 2.0, comparable to frontier models such as GPT-5.2 xHigh and Gemini 3.0 Pro. By redefining the efficiency frontier, Step 3.5 Flash provides a high-density foundation for deploying sophisticated agents in real-world industrial environments.

cs.CL↗

arXiv2026

Robust Grid-Forming Control Based on Virtual Flux Observer

This paper investigates the design and analysis of a novel grid-forming (GFM) control method for grid-connected converters (GCCs). The core novelty lies in a virtual flux observer-based synchronization and load angle control method. The terminal voltage of the converter is directly regulated to provide voltage-source behavior. The control parameters are designed for decoupling and pole placement. The proposed method exhibits strong robustness in stability and dynamical performance across varying and uncertain grid strengths. The robust control performance of the proposed method is first demonstrated by small-signal analysis, then validated by experiments on a 20 kVA power conversion system.

eess.SY↗

arXiv2026

VulnResolver: A Hybrid Agent Framework for LLM-Based Automated Vulnerability Issue Resolution

As software systems grow in complexity, security vulnerabilities have become increasingly prevalent, posing serious risks and economic costs. Although automated detection tools such as fuzzers have advanced considerably, effective resolution still often depends on human expertise. Existing automated vulnerability repair (AVR) methods rely heavily on manually provided annotations (e.g., fault locations or CWE labels), which are often difficult and time-consuming to obtain, while overlooking the rich, naturally embedded semantic context found in issue reports from developers. In this paper, we present VulnResolver, the first LLM-based hybrid agent framework for automated vulnerability issue resolution. VulnResolver unites the adaptability of autonomous agents with the stability of workflow-guided repair through two specialized agents. The Context Pre-Collection Agent (CPCAgent) adaptively explores the repository to gather dependency and contextual information, while the Safety Property Analysis Agent (SPAAgent) generates and validates the safety properties violated by vulnerabilities. Together, these agents produce structured analyses that enrich the original issue reports, enabling more accurate vulnerability localization and patch generation. Evaluations on the SEC-bench benchmark show that VulnResolver resolves 75% of issues on SEC-bench Lite, achieving the best resolution performance. On SEC-bench Full, VulnResolver also significantly outperforms the strongest baseline, the agent-based OpenHands, confirming its effectiveness. Overall, VulnResolver delivers an adaptive and security-aware framework that advances end-to-end automated vulnerability issue resolution through workflow stability and the specialized agents' capabilities in contextual reasoning and property-based analysis.

cs.SE↗

arXiv2026

A Construction Framework of Coded Caching Scheme for Multi-Access MISO Systems via Knapsack Problem

This paper investigates the coded caching problem in a multi-access multiple-input single-output (MAMISO) network with the combinatorial topology. The considered system consists of a server containing $N$ files, $Λ$ cache nodes, and $K$ cache-less users, where each user can access a unique subset of $r$ cache nodes. The server is equipped with $L$ transmit antennas. Our objective is to design a caching scheme that simultaneously achieves a high sum Degree of Freedom (sum-DoF) and low subpacketization complexity. To address this challenge, we formulate the design of multi-antenna placement delivery arrays (MAPDA) as a $0$--$1$ knapsack problem to maximize the achievable DoF, thereby transforming the complex combinatorial caching structure into a tractable optimization framework that yields efficient cache placement and flexible delivery strategies. Theoretical and numerical analyses demonstrate that: for networks with combinatorial topologies, the proposed scheme achieves a higher sum-DoF than existing schemes. Under identical cache size constraints, the subpacketization level remains comparable to existing linear subpacketization schemes. Moreover, under specific system conditions, the proposed scheme attains the theoretical maximum sum-DoF of $\min\{L+KM/N, K\}$ while achieving further reductions subpacketization. For particular combinatorial structures, we further derive optimized constructions that achieve even higher sum-DoF with lower subpacketization. ```

cs.IT↗